Statutory Damages Under the Copyright Act: An Empirical Study

The Copyright Act contains a powerful set of remedies for copyright infringement, including injunctive relief, monetary remedies, and costs and attorney's fees. 1 The statute divides monetary damages into two categories. First, it permits all prevailing plaintiffs to recover actual damages that they have suffered as a result of the infringement, plus any related profits earned by the infringer. 2 Second, prevailing plaintiffs may elect to recover an award of statutory damages for each work infringed in lieu of actual damages and profits. 3 The Copyright Act's statutory damages range is wide. Although the ordinary minimum award is $750 per infringed work and the ordinary maximum is $30,000 per infringed work, awards as small as $200 and as large as $150,000 per infringed work may be made depending on whether the infringement is judged to be innocent or willful. 4 Moreover, other than specifying minimum and maximum amounts per work infringed, the statute sets out no criteria governing where in the range an award should fall in a particular case. The statute only instructs courts to award statutory damages within the prescribed range in an amount that "the court considers just." 5 In this Article, we are interested in understanding what courts do in statutory damages cases and whether, given the absence of statutory direction, there are any discernable patterns in statutory damages outcomes under the Copyright Act. To this end, we constructed a new dataset of copyright cases in which statutory damages were awarded by a jury or a judge on the merits from January 1, 2009, through May 31, 2020 (n=277 awards). 6 In addition to recording amounts awarded by the court and the number of works infringed in each case, we collected data on the type of works that had been infringed, the level of defendant culpability, whether the award was made by a judge or jury, any lost licensing fee evidence presented by the plaintiff, and any damages requests made by the plaintiff.

Overall, we find that awards spanned the entire statutory damages range during the period we examined. Most of the awards, however, were concentrated at the lower end of the damages range. The most common amount awarded, in fact, was the ordinary minimum of $750 per work, and more than half of the awards in our dataset were less than $6,000 per work. But large awards were by no means unheard of. Two of the five most common award amounts in our dataset were $30,000 per work and $150,000 per work. Nearly a fifth of the awards exceeded the ordinary maximum of $30,000 per work.

We find that the type of work at issue in a case helps to explain some of the variation in award outcomes. Awards in cases involving movies, photographs and images, printed materials, and the public performance of songs, for instance, were generally small, with medians between $1,214 and $4,063 per work. Awards in cases involving fashion designs, software and video games, and artwork and 5. 17 U.S.C. § 504(c). The Seventh Amendment provides a right to a jury trial if the copyright owner elects statutory damages, although in practice awards are often determined by judges. See Feltner v. Columbia Pictures Television, Inc., 523 U.S. 340, 355 (1998); see also infra note 49 and accompanying text.

6. See infra Section II.A. Our dataset is limited to awards made in contested cases and excludes 2,701 default judgment awards that were made between January 1, 2009, and May 31, 2020. illustrations, on the other hand, were typically much larger, with medians between $22,000 and $50,000 per work. Still, there was a great deal of variation within most of the categories we examined. The middle 50% of awards in cases involving printed materials, for example, ranged from $750 to $52,500 per work. Exceptions included cases involving software and video games and cases involving the public performance of songs. Awards in these two categories tended to fall within relatively narrow ranges.

Culpability is another key explanatory factor. Awards tended to be larger the more severe the infringement. The median award in cases involving willful infringement, for instance, was $10,000 per work compared to median awards of $200 and $3,000 per work in cases involving innocent infringement and nonwillful infringement, respectively. Awards for willful infringement, however, were highly variable. The middle 50% of awards for willful infringement ranged from $3,000 up to $60,000 per work. Awards for innocent and nonwillful infringement, on the other hand, were relatively concentrated. The middle 50% of awards for innocent infringement ranged from $200 to $698 per work, and the middle 50% of awards for nonwillful infringement ranged from $750 to $7,400 per work.

Whether a jury or a judge made the award also mattered. Overall, the median jury award was $20,000 per work, and the median judge award was $3,775 per work. The gap between jury awards and judge awards was largest when there was a finding of willful infringement. In these instances, the median jury award was $50,970 per work, and the median judge award was $24,856 per work. 7 Juries awarded enhanced damages greater than $30,000 per work 60% of the time when there was a finding of willful infringement compared to 35% of the time for judges. 8 Jury awards, however, were widely dispersed. The middle 50% of jury awards in willful infringement cases ranged from $9,250 to $100,000 per work, while the middle 50% of judge awards in willful infringement cases ranged from $3,000 to $25,000 per work.

Finally, we find that lost licensing fee evidence and plaintiffs' award requests, when presented, help to explain variation in award outcomes. On average, courts awarded between two and three times the lost licensing fee when the plaintiff presented lost licensing fee 7. Note that these medians exclude awards made in public performance of songs cases because all but one were made by judges.

8. Again, these calculations exclude awards made in public performance of songs cases. evidence. Moreover, courts awarded the full amount requested by the plaintiff 58% of the time when plaintiffs made a specific damages request. Damages requests were granted in full more often in cases involving certain types of works. In cases involving printed materials, software, and music, for example, damages requests were granted more than 70% of the time. In cases involving the public performance of songs, it was common for plaintiffs to both present lost licensing fee evidence and make a damages request. In these instances, courts tended to grant the full amount requested when the request was no greater than three times the value of the lost licensing fee.

Although the details of each copyright infringement case are different, and our dataset is limited to a relatively small number of awards in contested cases, we believe that the data and findings we present in this Article can provide useful context to attorneys who litigate copyright cases and judges and juries tasked with determining statutory damages awards under the Copyright Act. 9 We also believe 9. We realize that the unpredictability of damages is, in some cases, a feature of the system that benefits copyright plaintiffs: the possibility of an outsized statutory damages award, one which substantially exceeds any provable amount of actual damages and profits, is doubtless often used as leverage to increase the settlement value of infringement claims. But viewed from a more detached perspective, while compensation, deterrence, and even a measure of punishment in appropriate cases can all be defended as desirable features of copyright's damages regime, unpredictability, as such, cannot.

As Sam Bray explains, a benefit of statutory damages is their predictability. See Samuel L. Bray, Announcing Remedies, 97 CORNELL L. REV. 753, 756-57 (2012); see also Randall R. Bovbjerg et al., Valuing Life and Limb in Tort: Scheduling "Pain and Suffering," 83 NW. U. L. REV. 908, 925 (1989) ("[H]ighly variable, unpredictable valuations undercut the deterrence function of tort law."). Bray observes that recent criticisms of excessive statutory damages awards underestimate their "precommitment function." Bray, supra, at 756-757, 785 n.140. He points out that it can make sense to commit in advance to imposing massive penalties on sophisticated actors for the sake of deterring bad conduct if the "legal decision makers know the socially optimal behavior but do not know the external costs of nonoptimal behaviors." Id. at 756, 782. Sophisticated actors are likely to be aware of the penalty and adjust their behavior accordingly, for example, by adapting an effective compliance system. Id. But the Copyright Act contains what Bray labels an "intermediate approach" in which damages are tailored within a wide predetermined range. Id. at 794-95. As Bray recognizes, the precommitment rationale does not justify an intermediate approach since there is no guarantee that the damages imposed would be at the high end of the range. For unsophisticated actors who are likely to underestimate penalties, Bray also suggests that deterrence may be better achieved by announcing remedies in advance. Id. at 756, 776-81. But he also questions the usefulness of an intermediate approach in this context because unsophisticated actors will have trouble grasping where in the range the remedy is likely to fall. Id. at 795-96.

that our data and findings will be useful to the new Copyright Claims Board (CCB), established by the Copyright Alternative in Small-Claims Enforcement Act of 2020 (CASE Act). 10 The statute authorizes the CCB to award statutory damages for timely registered works up to $15,000 per work infringed, with the maximum statutory damages award for any one CCB proceeding set at $30,000. Additionally, for works that are not timely registered in accordance with the requirements of 17 U.S.C. § 412, the CASE Act authorizes the CCB to award statutory damages not exceeding $7,500 per work infringed, with a total of $15,000 in any one proceeding. 11 The availability of statutory damages for unregistered works, even though in a lower amount, carries the potential for a substantial expansion of the number of infringement cases in which statutory damages are awarded.

In light of the changes enabled by the CASE Act, and even though the maximum total award in any one CCB proceeding is lower than the maximum award for copyright infringement litigation in federal court, we hope that our results will help the CCB's decisionmakers, who are copyright experts, craft statutory damages awards that may serve as models to help guide generalist federal judges, both when they themselves make statutory damages awards and as they provide guidance to juries in rendering such awards.

This Article proceeds as follows. In Part I, we describe the Copyright Act's statutory damages provisions, detail the justifications that have been offered for the inclusion in the statute of statutory damages as a form of monetary relief, and summarize some of the problems that the Copyright Act's particular statutory damages provision creates for potential defendants attempting to gauge the likely consequences of infringement, as well as for courts and juries seeking to make appropriate money damages awards in copyright cases. In Part II, we present our empirical study of statutory damages awards in copyright infringement cases. Here, we describe our dataset and present detailed results from our analyses. In Part III, we consider the study's implications for setting appropriate statutory damages awards in copyright infringement cases. In particular, we argue that our empirical evidence strongly suggests that in copyright infringement cases in which information is available that permits the approximation of actual damages, courts tend to award statutory damages that are guided by, and are based on a reasonable multiple of, those approximated actual damages. Given the systemic concerns raised by the very wide possible range of statutory damages awards that the Copyright Act makes available, we argue that courts should encourage parties to offer evidence from which actual damages can be approximated, even if only roughly. Doing so is likely to result in a salutary shift that will both make statutory damages more predictable and better align those awards with the compensation, deterrence, and in appropriate cases, punishment goals of the Copyright Act's remedial regime.

A. The Rules Governing Statutory Damages Under 17 U.S.C. § 504 Congress has established an unusual system of remedies for copyright infringement. 12 Under § 504 of the Copyright Act, a victorious copyright plaintiff may choose between actual damages or, for plaintiffs that have timely registered their copyright claims, statutory damages. 13 A plaintiff seeking actual damages is "entitled to recover the actual damages suffered by him or her as a result of the infringement, and any profits of the infringer that are attributable to the infringement and are not taken into account in computing the actual damages." 14 The copyright owner must prove the actual damages resulting from the defendant's infringement by a For published works, statutory damages are available if the work is registered in advance of the infringement or if it is registered within three months after first publication. Id. § 412(2). Section 412 includes certain exceptions to these registration requirements-for example, they do not apply to "an action brought for a violation of the rights of the author under section 106A(a)" (i.e., for violation of the author's rights to attribution and integrity added to the Copyright Act by the Visual Artists Rights Act). Id. § 412. By making registration a prerequisite to the recovery of statutory damages, the Copyright Act encourages copyright owners to register their works promptly and also encourages would-be infringers to check to see whether a work has been registered. See Johnson v. Jones, 149 F.3d 494, 505 (6th Cir. 1998).

14. 17 U.S.C. § 504(b). As the legislative history explains, "Damages are awarded to compensate the copyright owner for his losses from the infringement, and profits are awarded to prevent the infringer from unfairly benefiting from his wrongful act." S. REP. NO. 94-473, at 143 (1976).

preponderance of the evidence. 15 For copyright claims that have been timely registered, the plaintiff may opt instead for statutory damages "at any time before final judgment is rendered." 16 For statutory damages, the copyright owner need not prove actual damages but is entitled to recover an amount within the statutory range established by Congress. 17 That range, as previously noted, is very wide. Section 504(c)(1) sets a statutory damages range of $750 to $30,000 per work infringed. 18 The statute also gives courts discretion to award enhanced damages up to $150,000 per work for willful infringement and reduced damages down to $200 per work for innocent infringement. 19 That is, statutory damages can range from as low as $200 per work infringed in cases of innocent infringement to as much as $150,000 per work infringed in cases of willful infringement, with an award ranging between $750 and $30,000 in the ordinary case. 20 These damages, which may be awarded without any proof that the plaintiff suffered any harm or that the defendant profited from the infringement, permit maximum awards for ordinary, nonwillful infringement up to 40 times the minimum award, and maximum awards for willful infringement up to 200 times the minimum. As 15. See, e.g., Smith v. Thomas, 911 F.3d 378, 381 (6th Cir. 2018). Note, however, that the Copyright Act provides substantial aid to the plaintiff in proving defendant's profits-the plaintiff is required only to introduce evidence of the infringer's gross revenues related to the infringement, and then the burden shifts to the infringer to prove both the expenses that should be deducted from gross revenues to calculate profits and also the elements of profit that are not attributable to infringement (e.g., profits due to the defendant's own non-infringing creative contributions or business acumen). See 17 U.S.C. § 504(b) ("In establishing the infringer's profits, the copyright owner is required to present proof only of the infringer's gross revenue, and the infringer is required to prove his or her deductible expenses and the elements of profit attributable to factors other than the copyrighted work.").

16. 17 U.S.C. § 504(c)(1). 17. 20. Id. § 504(c). As mentioned earlier, for CASE Act proceedings brought before the Copyright Claims Board, works registered in accordance with § 412 are eligible for up to $15,000 in statutory damages per work infringed, with a total award of not more than $30,000 in a single CCB proceeding. Id. § 1504(e)(1). For works that are not timely registered in accordance with § 412, statutory damages may not exceed $7,500 per work infringed, with a total of $15,000 in any one proceeding. Id. § 1504(e)(1)(A)(ii)(II). many have observed, the breadth of the Copyright Act's potential range of statutory damages poses difficulties for defendants accused of copyright infringement in attempting to gauge the likely size of their exposure. 21 Recent empirical research shows that the unpredictability of statutory damages allows plaintiffs to induce riskaverse defendants to settle, even in cases in which they should prevail. 22 This difficulty is magnified by the absence of rules (either in the statute itself or developed by courts) for determining where in that enormous range a particular damages award should fall. Plaintiffs seeking statutory damages ordinarily are not required to offer evidence establishing their actual damages or defendant's profits from infringement, and in fact, some courts have held that an award of statutory damages has no necessary connection to the damages a plaintiff actually suffered. 23 Courts and juries are thus left at large to apply their intuition, and this has resulted, as we shall see, in a wide 2020). While our focus in this Article is on statutory damages, risk-averse defendants must also be wary of § 504's provision allowing a copyright owner recovering actual damages to also recover the profits attributable to the infringement. For example, in the Google/Oracle litigation discussed below, Google's profits totaled more than ten times Oracle's claimed actual damages. ) (rejecting requested award on default judgment because the plaintiff "presents no evidence to which the Court might anchor such an award, for instance the amount of actual damages or the licensing fee for the Photograph"); H.R. REP. NO. 94-1476, at 161 (1976) (explaining that "there is nothing in section 504 to prevent a court from taking account of evidence concerning actual damages and profits in making an award of statutory damages within the range set out in subsection (c)"). range of awards even within categories of cases that otherwise appear alike.

To fully understand the costs of this unpredictability, it is necessary to realize that the optimal amount of unauthorized copying is not zero. The conduct at issue in copyright infringement lawsuitsunauthorized copying, distribution, adaptation, public performance, or display-is not inherently wrong: copying is not, in the lawyers' argot, "malum in se." 24 We condemn certain forms of copying as copyright infringement for instrumental reasons-i.e., because we fear that it will depress authors' incentives to create. 25 But there are many instances in which we permit copying-even unauthorized copyingbecause we believe it to be socially productive. That is why the Copyright Act contains an extensive set of exceptions and limitations to copyright protection, including the well-known provision protecting fair use. 26 And that is also why the Copyright Act declares certain building-block elements of expression, including facts, ideas, principles, and processes, to be outside the scope of copyright entirely. 27 Anyone may copy these elements at will; a copyist cannot be held liable, for example, for copying names and telephone numbers from a telephone directory, even when the purpose of doing so is to market a competing directory. That result, the Supreme Court held in Feist Publications, Inc. v. Rural Telephone Service Co., is "neither unfair nor unfortunate. It is the means by which copyright advances the progress of science and art." 28 Given that copying is not inevitably socially harmful, and indeed is in many cases socially beneficial, the uncertainty created by the Copyright Act's statutory damages rules may deter otherwise socially productive behavior. These statutory damages rules may over-deter if they discourage behavior that, viewed prospectively, may be lawful under a copyright exception. By exposing copyists to the possibility of a very large statutory damages award, even for these close cases, we discourage parties from engaging in conduct that tests the boundaries of copyright protection. We also discourage those who engage in such potentially lawful conduct and are threatened with a copyright infringement lawsuit from standing firm and litigating their defense through to a conclusion. In the process, we lose information that might help us better place the boundary between infringing, socially harmful uses and protected, socially valuable usesinformation that would reduce uncertainty and lower the cost of engaging in the socially productive copying that copyright law aims to promote. 29 The likelihood that copyright law's statutory damages rules may over-deter is increased by the fact that the border between infringing and non-infringing conduct can be difficult to discern in advance. The Copyright Act's fair use provision, for example, is framed as a flexible standard, rather than as a set of rules drawn with precision, and so in many cases it will not be clear whether particular copying is fair use or infringement until a court rules on the question. 30 The same is true 29. Cf. Christopher Sprigman, Copyright and The Rule of Reason, 7 J. TELECOMM. & HIGH TECH. L. 317, 323-24 (2009) (arguing that in some copyright infringement cases placing the burden of proof on the plaintiff to show that infringement caused actual harm would produce evidence about harmful versus harmless forms of infringement that current law is not structured to provide).

30. The Copyright Act's fair use provision, characterized in the Act's legislative history as an "equitable rule of reason," H.R. REP. NO. 94-1476, at 65 (1976), directs courts to balance at least four relevant factors:

Notwithstanding the provisions of sections 106 and 106A, the fair use of a copyrighted work, including such use by reproduction in copies or phonorecords or by any other means specified by that section, for purposes such as criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copyright. In determining whether the use made of a work in any particular case is a fair use the factors to be considered shall include-(1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

(3) the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and (4) the effect of the use upon the potential market for or value of the copyrighted work. The fact that a work is unpublished shall not itself bar a finding of fair use if such finding is made upon consideration of all the above factors. 17 U.S.C. § 107. Because our dataset consists only of cases in which the court awarded statutory damages, it does not contain any cases in which the defendant prevailed on a fair use defense. Other scholars have done empirical work on fair use defenses. See Matthew Sag, Predicting Fair Use, 73 OHIO ST. L. J. 47, 49 (2012); Neil Weinstock of the statute's distinction between uncopyrightable ideas and copyrightable expression. 31 Whether copying in a particular case is shielded by the § 102(b) exclusions often is as unclear ex ante as whether copying qualifies as fair use.

A recent case provides an apt example of the vagaries of both fair use and the boundary between unprotected ideas and protected expression. 32 In 2010, Oracle sued Google for using portions of Oracle's Java "declaring" code in Google's Android platform for mobile telephones. 33 That case presented the issue of whether the Java declarations, which functioned merely as labels for certain operations rather than as code implementing the operations themselves, should be classed as "methods of operation" that § 102(b) of the Copyright Act makes uncopyrightable. 34 After a district court held that the declaring code was uncopyrightable, the Federal Circuit, purporting to apply Ninth Circuit law, held otherwise. 35 After remand and a trial, a jury found for Google, holding that the company's use of Java declarations was fair use. 36 The Federal Circuit then reversed the jury's verdict, holding as a matter of law that Google's use was not fair. 37 The Supreme Court reversed that ruling, holding that Google's use of the Java declarations was a fair use as a matter of law, and in the process, cast substantial doubt on the Federal Circuit's holding on the copyrightability question. 38 The history of this case illustrates the difficulty that potential defendants may face in knowing in advance whether their conduct is infringing. Both the copyrightability of the Java declarations, and if those declarations were copyrightable, whether Google's use was fair, seesawed back and forth as the case made its way up and down the judicial hierarchy. And because Google's use of the Java declarations to build a new mobile telephone operating system potentially involves socially productive conduct that we wish to encourage (the Supreme Court held that it did), the Google v. Oracle dispute stands for a much broader problem: We should be wary of exposing defendants whose copying may be socially productive to excessive damages for ending up on the wrong side of an indistinct line. 39 In addition, once potential damages grow large enough, we should worry about excessively variable damages. Both can discourage defendants from undertaking socially productive activities for fear that they will eventually be found to have strayed over an indistinct border into potentially crushing liability.

38. See Google LLC v. Oracle Am., Inc., 141 S. Ct. 1183, 1190 (2021); id. at 1202 ("Although copyrights protect many different kinds of writing . . . we have emphasized the need to 'recogni[ze] that some works are closer to the core of [copyright] than others.' In our view, for the reasons just described, the declaring code is, if copyrightable at all, further than are most computer programs (such as the implementing code) from the core of copyright." (quoting Campbell v. Acuff-Rose Music, 510 U.S. 569, 586 (1994))); see also id. at 1201 ("The declaring code at issue here resembles other copyrighted works in that it is part of a computer program. Congress has specified that computer programs are subjects of copyright. It differs, however, from many other kinds of copyrightable computer code. It is inextricably bound together with a general system, the division of computing tasks, that no one claims is a proper subject of copyright. It is inextricably bound up with the idea of organizing tasks into what we have called cabinets, drawers, and files, an idea that is also not copyrightable. It is inextricably bound up with the use of specific commands known to programmers, known here as method calls (such as java.lang.Math.max, etc.), that Oracle does not here contest.").

39. See, e.g., id. at 1208 ("[G]iven programmers' investment in learning the Sun Java API, to allow enforcement of Oracle's copyright here would risk harm to the public. Given the costs and difficulties of producing alternative APIs with similar appeal to programmers, allowing enforcement here would make of the Sun Java API's declaring code a lock limiting the future creativity of new programs. Oracle alone would hold the key.").

Although Congress did not specify an approach to shaping appropriate statutory damages awards in particular cases, this does not mean that Congress was indifferent to the question. Congress wrote into the Copyright Act a directive that statutory damages should be set "as the court considers just." 40 Within the broad parameters set out in the statute, Congress sought "to provide the courts with reasonable latitude to adjust recovery to the circumstances of the case, thus avoiding some of the artificial or overly technical awards" that had resulted under existing law. 41 The question, however, is how to operationalize Congress's intent in copyright infringement litigation.

The federal appellate courts have provided trial courts with only general guidelines for determining a just award. The Second Circuit, for instance, sets forth six factors: (1) the infringer's state of mind; (2) the expenses saved and profits earned by the infringer; (3) the revenue lost by the copyright holder; (4) the deterrent effect on the infringer and third parties; (5) the infringer's cooperation in providing evidence concerning the value of the infringing material; and (6) the conduct and attitude of the parties. 42 Other circuits use similar factors. 43 Given these broad standards and the extremely deferential standard of review on appeal, some appellate courts have concluded that the district courts enjoy "almost unfettered discretion" in setting awards of statutory damages. 44 The Seventh Amendment adds a further wrinkle. While § 504's reference to "the court" seems to reflect Congress's expectation that statutory damages determinations would be made by judges, the 40. 17 U.S.C. § 504(c)(1). 41. H.R. REP. NO. 94-1476, at 161 (1976). 42. Bryant v. Media Right Prods., Inc., 603 F.3d 135, 144 (2d Cir. 2010). 43. See, e.g., Chi-Boy Music v. Charlie Club, Inc., 930 F.2d 1224, 1229 (7th Cir. 1991) ("The court was not required to follow any rigid formula. Indeed, district courts enjoy wide discretion in awarding fees and may consider various factors such as 'the difficulty or impossibility of proving actual damages, the circumstances of the infringement, and the efficacy of the damages as a deterrent to future copyright infringement.'"); Broad. Music, Inc. v. Evie's Tavern Ellenton, Inc., 772 F.3d 1254, 1261 (11th Cir. 2014) ("In calculating damages, courts generally consider: (1) the infringers' blameworthiness (willful, knowing, or innocent); (2) the expenses saved and the profits reaped by the defendants in connection with the infringement; (3) the revenues lost by the plaintiffs due to the defendants' conduct; and (4) the deterrent value of the damages imposed.").

44. See Broad. Music, Inc. v. Star Amusements, Inc., 44 F.3d 485, 489 (7th Cir. 1995); Cullum v. Diamond A Hunting, Inc., 484 F. App'x 1000, 1002 (5th Cir. 2012).

Supreme Court held subsequent to the passage of the Copyright Act that the Seventh Amendment guarantees a jury trial "on all issues pertinent to an award of statutory damages under § 504(c) of the Copyright Act, including the amount itself." 45 Juries likewise enjoy wide latitude in assessing statutory damages. 46 The judge will instruct the jury using the factors mentioned above, but the jury must ultimately decide how to apply these tests on a case-by-case basis. One leading treatise even claims that "it is doubtful that juries can be meaningfully instructed to compare the facts at bar against those of prior cases in order to slot an appropriate award into the scheme of precedent." 47 For reasons we will explain later, we do not agree that juries cannot be meaningfully instructed on statutory damages. We believe there are ways that judges can help juries to formulate more just and predictable statutory damages awards, without displacing the jury's ultimate authority to determine the size of such awards. 48 Note that the Seventh Amendment does not keep all statutory damages decisions out of the hands of judges. For one thing, judges often decide statutory damages questions on summary judgment or in bench trials without the help of the jury. 49 While the statute and case law are generally open-ended, some judges have applied a rough rule of thumb that statutory damages awards generally should not exceed "a single-digit multiple of a reasonable licensing fee"-usually three 45. Feltner v. Columbia Pictures Television, Inc., 523 U.S. 340, 355 (1998). 46. Jury instructions, for example, simply reflect the open-ended case law developed by the courts. For example, the Seventh Circuit lists the following factors: 1) the expenses that Defendant saved and the profits that he earned because of the infringement; 2) the revenues that Plaintiff lost because of the infringement; 3) the difficulty of proving Plaintiff's actual damages; 4) the circumstances of the infringement; 5) whether Defendant intentionally infringed Plaintiff's copyright; and 6) deterrence of future infringement. COMM. ON that Feltner "does not mean, however, that a jury must resolve every dispute. When there are no disputes of material fact, the court may enter summary judgment without transgressing the Constitution"). Bench trials are also a possibility if the parties forgo their right to a jury trial.

to five times the lost licensing fee. 50 But judges have applied this framework only to the subset of cases in which lost licensing fees are known, and nothing requires a copyright owner electing statutory damages to introduce evidence of lost licensing fees or defendant's profits. 51 It is also important to note that the Copyright Act does very little to clarify when infringement qualifies as willful. 52 While there is legislative history suggesting that Congress intended enhanced damages only in "exceptional cases," courts have been more willing to find willfulness, holding that enhanced statutory damages are available when the plaintiff establishes either that the infringement was knowing or that the defendants acted in reckless disregard of the possibility that their actions infringed the owner's copyright. 53 As we shall see, the factors that courts have articulated are relevant to the determination of a proper statutory damages award. 54 But these factors are not particularly constraining: courts exercise 51. See 5 NIMMER, supra note 17, § 14.04(A). 52. See 17 U.S.C. § 504(c)(3)(A) (creating a rebuttable presumption that an infringer who provides false contact information to a domain name registry acts willfully but not otherwise defining willfulness). Under § 504(c)(2), innocent infringement occurs when the "infringer was not aware and had no reason to believe that his or her acts constituted an infringement of copyright." Id. § 504(c)(2). But reduced damages for innocent infringement are rare. almost complete discretion in setting awards, as do juries. The result is that copyright plaintiffs wield a powerful weapon of uncertainty, which in the words of one commentator, permits many copyright plaintiffs to "leverage the risk aversion of defendants to induce generous settlement concessions." 55

Several commentators have proposed reforms to the Copyright Act's statutory damages scheme. In general, these reforms are designed to provide copyright defendants with greater ability to assess in advance of litigation their likely exposure, either by lowering the ceiling for statutory damages or by imposing standards governing statutory damages awards. 56 But Congress has shown little appetite for altering the Copyright Act's statutory damages rules-in fact, when Congress has revised statutory damages, it has been to increase the amounts recoverable without changing the basic scheme. 57 Courts have resisted arguments from copyright scholars that due process limitations on punitive damages should limit jury awards of statutory damages falling within the statutory range. 58 or reduce jury awards within the statutory range, even if they are disproportionate to actual damages. 59 Nevertheless, as noted, Congress did recently establish a "small claims" system administered by the United States Copyright Office for copyright owners seeking less than $30,000. 60 This system caps statutory damages at $15,000 per work infringed for works timely registered-with a total award not to exceed $30,000. Notably, the CASE Act permits an award of reduced statutory damages-$7,500 per work, with the total award limited to $15,000-for works that do not meet the timely registration requirement that the Copyright Act imposes as a prerequisite for statutory damages generally. 61 But participation is voluntary, and both must agree to participate in proceedings before the CCB.

As a preliminary matter, one might ask what role statutory damages play in the Copyright Act's remedial scheme in light of the statute's robust rule for actual damages and disgorgement. That rule, which permits successful plaintiffs to recover both their actual damages and defendants' profits, means that an actual damages award is not limited to compensating the plaintiff. The prospect that an infringer will be forced to disgorge its profits attributable to infringement provides significant deterrence, and that deterrence is bolstered further by the prospect, unusual in American law, that the defendant may be ordered to pay the successful plaintiff's attorney's fees and other costs of litigating. 62 59. See 61. 17 U.S.C. § 1504(e)(1)(A)(ii). 62. See id. § 505 ("In any civil action under this title, the court in its discretion may allow the recovery of full costs by or against any party other than the United States or an officer thereof. Except as otherwise provided by this title, the court may also award a reasonable attorney's fee to the prevailing party as part of the costs."). Indeed, the U.S. Copyright Act's statutory damages provisions are unusual when compared with the copyright laws of our trading partners. While copyright laws in most countries allow plaintiffs to be compensated for the actual damages attributable to infringement, and many countries' copyright laws also allow the disgorgement of defendants' profits attributable to infringement, comparatively few countries provide statutory damage awards for copyright infringement. See Pamela Samuelson et al., Statutory damages have several justifications. First, they are said to be necessary because of evidentiary limitations that may make it difficult to determine actual damages in copyright cases. 63 For instance, it may not be easy to determine how many times users downloaded a song that a defendant uploaded to a peer-to-peer filesharing system. 64 Second, to the extent statutory damages exceed actual damages, they also deter further copyright violations by the infringer and by others. 65 Such deterrence is thought to be necessary because it may not be cost effective for owners to vindicate their rights in individual copyright cases, especially in a digital age in which it is easy and inexpensive to reproduce a copyrighted work. As Roger D. Blair and Thomas F. Cotter have suggested, "Perhaps only the threat of a statutory damages award, which may be many times greater than the actual harm or benefit derived from the defendant's unauthorized use, will be sufficient to prevent the value of the owner's copyright from being destroyed by a multitude of small-scale infringing acts." 66 In sum, statutory damages are said to provide a way to make copyright owners whole and to deter would-be infringers given the unique nature of copyright infringement.

But many scholars find these rationales unconvincing. 67 For one thing, conventional explanations do not seem to match what courts are actually doing. In practice, for instance, courts do not seem to take into account the costs of underenforcement in setting statutory damages. 68 Sometimes, they opt for a mere "mechanical formula of multiplication" by a single digit figure in cases in which there is evidence that makes actual damages calculable. 69 In many other cases, as our data shows, courts are following no discernable framework or pattern where actual damages are unknown. 70 Nor do courts seem to be compensating, at least not explicitly, for evidentiary limitations in establishing an owner's actual damages-there is no obvious link discernable in the cases between proof difficulties and statutory damages awards. 71 Given that statutory reform seems unlikely, we gathered empirical data on statutory damages in an effort to identify and explain variation in award outcomes under the current statutory and doctrinal framework. We believe that such context could help litigants, courts, and juries (under courts' tutelage) achieve more consistent and appropriately sized statutory damages awards without significant doctrinal innovation. As we discuss below, lawyers for copyright owners can introduce, and judges can consider, evidence that can be used to approximate actual damages. And providing such information to juries may allow courts to instruct juries regarding the relationship in prior cases between approximated actual damages and statutory damages awards, even if the jury is under no obligation to treat such information as dispositive. on the merits from January 1, 2009, through May 31, 2020. 72 Our dataset includes 277 statutory damages awards made in 240 cases. 73 For each award, we identified the total amount of the statutory damages awarded and the number of works infringed. From this information, we calculated the amount awarded per work infringed. In addition, we recorded whether the damages were decided by a jury or a judge (on summary judgment or in a bench trial) and whether there was a finding of willful or innocent infringement that allowed for enhanced or reduced damages. We also registered the amount of any lost licensing fees presented to the court, noted if the plaintiff requested a particular award amount, and in cases decided by a judge, documented any express rationale provided by the court for the award.

After gathering this information, we classified awards by the type of work that was infringed. These classifications include artwork and illustrations; fashion designs; other types of designs not related to 72. See District Court Overview, LEX MACHINA, https://law.lexmachina.com/help/documentation/district-court/overview. [https://perma.cc/8GTE-WCAX] (last visited Apr. 17, 2023). Our dataset excludes default judgments. According to the Lex Machina legal analytics database, there were 2,701 default judgment awards of statutory damages in copyright cases between January 1, 2009, and May 31, 2020. We have excluded default judgments on the assumption that the effort and expense required to obtain damages on default judgment makes these cases less comparable to cases decided by a judge or jury on the merits. But given the high number of statutory damages cases involving default judgments, this is an area that would benefit from future research.

73. See id. Our primary units of analysis in this Article are awards, not cases. The number of awards is greater than the number of cases because multiple awards are sometimes made in the same case, either because there are multiple parties or because different works have been infringed in different ways. To construct the dataset, we began by searching the Lex Machina legal analytics database using the following terms: "damage awards; with Copyright: Statutory Damages as a damage type; awarded on Judgment on Merits or Jury Verdict." We then examined the underlying docket entries for each award listed in the Lex Machina list and manually compiled information about each, such as the type of works infringed, the number of works infringed, and the amount sought by plaintiffs. In the process of reviewing the docket entries, we determined that it was appropriate to split four of the awards listed in the Lex Machina database into multiple awards. We also excluded 37 of the awards in the Lex Machina list from our analyses. An award was excluded from our analysis if: (a) it was vacated on appeal prior to May 31, 2020 (n=16); (b) critical information about the award, such as the number of works infringed, could not be ascertained from the docket entry (n=6); (c) our review of the docket entry indicated that the court calculated the award without taking into account the number of infringed works (n=5); (d) the award was the result of a default judgement (n=4); (e) the award was not made under § 504 of the Copyright Act (n=3); (f) the award listed in Lex Machina was actually an attorney's fee (n=2); and (g) damages were stipulated by the parties rather than the court (n=1). fashion; movies; music; photographs and images; printed materials; public performance of songs; software and video games; and television rights. 74 We also separated judge cases from jury cases and classified the awards by whether infringement was willful, nonwillful, or innocent. Table 1 shows the number of awards in each of these subgroups. The most common awards in our dataset were associated with the public performance of songs (n=87), photographs and images (n=36), music (n=31), fashion designs (n=28), and other designs unrelated to fashion (n=28). In combination, these five subject categories account for 76% of the 277 awards in the dataset. The dataset also includes more than twice as many statutory damages awards granted by judges through summary judgment or bench trials (n=200) than awards granted by juries (n=77). Disaggregating by culpability, 90% of the awards are concentrated in the judge/willful (n=101), the judge/nonwillful (n=93), and the jury/willful (n=54) categories.

74. The fashion designs category includes cases involving clothing, textile designs, handbags, jewelry, and tattoos. The other designs category includes cases involving architectural plans, labels, logos, packaging, toys, and other types of designs unrelated to fashion. We created separate categories for public performance of songs cases and television rights cases because they tended to have unique features that distinguish them from other music and movie cases. We placed each case in only one category-cases in the public performance of songs category were not also included in the music category, which involved other types of infringement, such as file sharing. B. Results

The 277 awards in our dataset were made by sixty-six U.S. district courts. The Southern District of New York (thirty-eight awards) and the Central District of California (thirty-seven awards) made the most statutory damages awards between January 1, 2009, and May 31, 2020. These courts were followed by the Middle District of Florida (twelve awards), the Northern District of Texas (ten awards), the Southern District of Texas (ten awards), the District of Massachusetts (nine awards), the Western District of Texas (nine awards), the District of Kansas (nine awards), and the Northern District of Ohio (seven awards). In sum, nine courts made over half the awards in our dataset. Most district courts decided very few copyright cases on the merits. Specifically, twelve courts made two awards, twenty-two courts made one award, and twenty-eight courts made no statutory damages awards between January 1, 2009, and May 31, 2020.

The awards in our dataset came in the form of seventy-nine distinct dollar amounts, all falling within the $200-$150,000 range set out in 17 U.S.C. § 504(c). Sixty percent of the awards in the dataset were comprised in a total of fourteen distinct dollar amounts. Common award amounts were generally round numbers. They also often involved key amounts set out in the statute, such as $200, $750, $30,000, and $150,000. The most common award, in fact, was the ordinary minimum of $750 per work. Courts awarded $750 in thirtyseven of 277 instances, or 13.4% of the time. The second most common amount was $3,000 per work, which courts awarded in eighteen instances (6.5%). Other common awards spanned the range allowed by the statute. Courts awarded $30,000 per work in fifteen instances (5.4%); $5,000 in fourteen instances (5.1%); $150,000 in twelve instances (4.3%); $1,000 and $100,000, each in ten instances (3.6%); $7,500 and $10,000, each in nine instances (3.3%); $2,000 in eight instances (2.9%); $200 in seven instances (2.5%); and $8,000, $9,250, and $15,000, each in six instances (2.2%). The remaining 40% of awards came in the form of sixty-five distinct dollar amounts, fortyseven of which only appear in the dataset one time.

Figure 1 shows the distribution of the 277 awards across the entire statutory damages range. Awards have been grouped into $6,000 increments. The histogram bars represent the percentage of the 277 awards in each $6,000 increment. Immediately, we notice that the majority of awards are concentrated at the lower end of the statutory damages range. The first bar from the left indicates, for instance, that 52% of the awards were less than or equal to $6,000 per work. 75 The remaining 48% of awards were dispersed between $6,000 and $150,000 per work. Courts awarded enhanced damages greater than $30,000 per work 18% of the time. Although enhanced damages are not the norm, 18% is not insignificant.

75. The median award was $5,874 per work and the mean was $23,848 per work. The mean is much larger than the median because of a relatively small number of very large awards on the upper end of the distribution. Because the median is not sensitive to outliers like the mean is, we believe it better describes the central tendency of this data.

Most awards in the dataset involved a small number of works. Of 277 awards, seventy-five involved just one infringed work, and 145 awards involved two to twelve works. Only thirteen of the 277 awards in our dataset involved more than 100 works, four of which involved more than 1,000 works. In one atypical case, a jury found that 10,017 works had been infringed. 76 76.

Figure 2 shows the number of awards (bars) and the median award per work (line) by the number of works infringed. Here, the data suggests that the amount awarded per work tended to decrease as the number of works at issue increased. The median award, for instance, was $10,000 when just one work was involved. The median award per work decreased to $9,625 when two works were involved, to $5,000 when three works were involved, and to $3,500 when four works were involved. After rising to $6,000 and $7,500 in instances involving five and six works, the median award decreased further to $750 per work in cases involving seven works and ranged from $2,500 to $3,000 thereafter. Seventy-nine percent of awards in our dataset involved twelve works or less. Beyond twelve works, it is not possible to identify trends between the median award per work and the number of works infringed due to the small number of observations in our dataset. Among the fifty-seven awards where more than twelve infringed works were involved, however, the median award was $3,000 per work.

Next, we group the awards by the type of work that was infringed. To compare the distribution of amounts for each category, Figure 3 shows box and whisker plots. The left hinge of the box represents the 25th percentile award, and the right hinge represents the 75th percentile award. The line in the middle of the boxes represents the 50th percentile, or median. The boxes thus represent the range of awards for the middle 50% of cases in each subject category. This range is a measure of statistical dispersion called the interquartile range, or IQR. A shorter box (i.e., a smaller IQR) means that the award amounts for a particular category were more concentrated. A longer box (i.e., a larger IQR) means that the award amounts were more widely dispersed or spread out across the distribution. Put another way, a shorter box signals that award amounts in a given category are less variable and thus more predictable; a longer box signals that awards are more variable and thus less predictable.

Other indicators of dispersion are the lines that emerge from the boxes, which are called whiskers. The whiskers often extend to the minimum and maximum award amounts in each category. Longer whiskers suggest greater variability between the 75th percentile and the maximum value and thus less predictability at the upper end of the distribution. The whiskers do not extend to the maximum value in categories where there are extreme outliers. 77 Extreme outlying values are represented by dots that lie to the right of the upper whisker. The rightmost dot is the maximum award amount in that category. Below the box and whisker plots, we include a table that lists the median, mean, 25th percentile, 75th percentile, minimum value, maximum value, and percentage of enhanced damages awards above $30,000 in each category. 78 77. We consider an outlier any award amount that is greater than the sum of the 75th percentile and one-and-a-half times the size of the interquartile range.

78. Corresponding with our focus on the median as the primary measure of central tendency in this Article, we discuss the spread of distributions in terms of ranges and the IQR rather than variance and standard deviation because variance and standard deviation are calculated using means and are thus sensitive to the outliers in many categories. Figure 3 shows that median awards were within the ordinary statutory damages range of $750 to $30,000 per work in nine of the ten categories. Disputes involving printed materials (n=24), public performance of songs (n=87), photographs and images (n=36), and movies (n=14) had the lowest median awards at $1,214, $3,000, $3,100, and $4,063 per work, respectively.

The awards for public performance of songs stand out not only because of their relatively low median, but also because the distribution of award amounts was so narrowly concentrated. The middle 50% of awards in this category ranged from $2,000 to $6,000-a difference, or IQR, of just $4,000. Because the distribution of awards was so concentrated, public performance of songs is the most predictable category in our dataset, a matter we will explore in detail later. 79 Public performance of songs awards are also unique because they never exceeded the ordinary maximum of $30,000 during the years we examined. The highest public performance of songs award in our dataset was $25,000 per work. 80 Awards in the movies, photos and images, and printed materials categories were more variable than awards in public performance of songs cases. The middle 50% of awards in the photos and images category, for instance, ranged from $750 to $20,000 per work (IQR=$19,250). For statutory damages involving movies, the middle 50% of awards ranged from $750 to $22,500 (IQR=$21,750). The printed materials category, despite having the lowest median award, was highly variable. Specifically, the middle 50% of printed materials awards ranged from $750 to $52,500 (IQR=$51,750).

The plots indicate that movies, photos and images, and printed materials awards were concentrated at the lower end of the awards range and became increasingly dispersed at the upper end of the awards range. 81 Photos and images awards, for instance, reached up to $150,000 per work, yet 28% of awards in this category were $750 per work or less, and 50% of awards were $3,100 per work or less. Thus, although large awards were made in categories such as photos, movies, and printed materials, smaller awards were more common. That said, many awards exceeded the ordinary maximum of $30,000. 81. This is evidenced by the small distance between the minimum value and the median compared to the much larger distance between the median and the rightmost whisker.

Twenty-five percent of printed materials awards, 22% of photos and images awards, and 14% of movies awards were greater than $30,000 per work.

Median awards for designs unrelated to fashion (n=28), music (n=31), and television rights (n=7) were higher at $7,576, $9,250, and $15,000 per work, respectively. Moreover, design, music, and television rights awards were similarly dispersed across the statutory damages range. The middle 50% of music awards ranged from $750 to $30,000 (IQR=$29,250), the middle 50% of design awards ranged from $5,000 to $40,000 (IQR=$35,000), and the middle 50% of television rights awards ranged from $8,900 to $44,500 (IQR=$35,600). Twenty-nine percent of television rights awards, 25% of design awards, and 16% of music awards exceeded $30,000 per work.

The three remaining categories had the highest median awards. At $22,000, the median award in fashion design cases (n=28) was large compared to the other categories discussed so far. Awards in this category were also highly dispersed. The middle 50% of awards in the fashion design category ranged from $8,750 to $61,667 per work (IQR=$52,917). Thirty-two percent of fashion design awards exceeded $30,000. The median software and video games award (n=11) was $30,000 per work. Although not as predictable as public performance cases, software and video games awards were quite concentrated relative to other categories. The middle 50% of software and video games awards ranged from $30,000 to $40,300 (IQR=$10,300). Thirty-six percent of software and video games awards exceeded $30,000. Finally, cases involving artwork and illustrations (n=11) had the highest median award at $50,000. Awards in this area were also the most widely dispersed, with the middle 50% of awards ranging from $18,852 to $95,000 per work (IQR=$76,148). Artwork and illustrations cases also had the highest share of enhanced damages awards. Sixty-four percent of artwork and illustrations awards exceeded $30,000 per work.

Figure 4 groups award amounts by whether infringement was innocent, nonwillful, or willful. As expected, awards were higher and more dispersed as the severity of infringement increased. Recall that although the normal range of statutory damages is between $750 and $30,000, § 504(c) allows for awards as low as $200 if infringement is considered innocent and as high as $150,000 if it is considered willful.

Figure 4 shows that awards were smallest when infringement was considered innocent (n=12). Seven of twelve innocent infringement awards were $200 per work, and all but one innocent infringement award was $1,000 per work or less. The outlier in the innocent infringement category was a software case that resulted in an award of $30,000 per work. 82 In cases where infringement was considered nonwillful, the median award amount was $3,000 per work, and the middle 50% of awards ranged from $750 to $7,400 (IQR=$6,650). Thirty percent of awards in nonwillful infringement cases were exactly the statutory minimum of $750 per work and 75% were $7,000 or less. Nine of 110 awards in the nonwillful category were the statutory maximum of $30,000. Three of the maximum awards were related to software, and 82. See Jury Verdict, at 4, 6-7, Oracle USA, Inc. v. Rimini St., Inc., No. 10-CV-0106-LRH-PAL, 2015 WL 7627428, at *3, *5 (D. Nev. Oct. 13, 2015), ECF No. 896 (awarding a high amount of $2.79 million for ninety-three infringed works (i.e., $30,000 per work) after the jury concluded that the fair market value of a license for the infringed works would have been $35.6 million). two were related to fashion. 83 The others involved logos, music, counterfeit textbooks, and illustrations. 84 Not surprisingly, a finding of willful infringement resulted in the highest awards (n=155). These were also the most widely dispersed awards. The median award in the willful infringement category was $10,000, and the middle 50% of awards ranged from $3,000 to $60,000 (IQR=$57,000). The most common award amounts in this category were $3,000 (thirteen awards), $150,000 (twelve awards), and $100,000 (ten awards). Courts awarded enhanced damages 32% of the time when there was a finding of willful infringement.

Figure 5 compares the distribution of awards granted by judges (n=200) with awards granted by juries (n=77). It is immediately apparent from this plot that judges tended to grant smaller awards than juries. First, observe that the median judge award of $3,775 was about one-fifth the size of the median jury award of $20,000. Second, while both judges and juries awarded damages as low as $200 and as high as $150,000, judge awards were concentrated at the lower end of that range. The middle 50% of judge awards ranged from $1,250 to $10,000-a difference of $8,750-and 88% of awards granted by judges were within the ordinary statutory damages range of $750 to $30,000 per work. Jury awards, in contrast, were widely dispersed.

The middle 50% of jury awards ranged from $6,000 all the way up to $99,830 per work-a difference of $93,830, or more than ten times the dispersion of judge awards. Third, judges were less likely to award enhanced damages or the statutory maximum. Only 9% of awards granted by judges exceeded $30,000 compared to 42% of awards granted by juries. Moreover, judges granted the maximum award of $150,000 only 2.5% of the time while juries granted the $150,000maximum 9% of the time. Figure 6 again compares judge and jury awards, but this time we disaggregate the awards by culpability. Whether culpability is innocent, nonwillful, or willful, we observe a similar pattern in which judge awards were smaller and more concentrated, and jury awards were larger and more widely dispersed. With regard to awards for innocent infringement, for instance, judge awards (n=6) were always $200 or $250 per work, with a median award of $200. Jury awards (n=6), on the other hand, ranged from $200 up to the $30,000 maximum and had a median of $698. With regard to nonwillful infringement, judges (n=93) and juries (n=17) had median award amounts of $2,750 and $5,000 per work, respectively. Juries and judges both awarded the minimum amount of $750 per work about 30% of the time. Jury awards in the nonwillful category, however, were somewhat more dispersed. Juries awarded the maximum amount of $30,000 about 12% of the time compared to about 8% of the time when awards were granted by judges. than judges to grant large awards and award enhanced damages when infringement was determined to be willful. Figure 6, however, likely overstates the differences between judge and jury awards because it does not take caseload composition into account. Recall from Figure 3 that awards in cases involving public performance of songs were small relative to other categories and concentrated at the lower end of the distribution. All but one of these small, concentrated awards were granted by judges. When we omit public performance awards from the analysis, the median jury award for willful infringement (n=54) remained $50,970, but the median judge award for willful infringement (n=52) increased from $7,500 to $24,856 per work. The distribution of judge/willful awards also became more dispersed when public performance awards were omitted, with the middle 50% of judge awards for willful infringement ranging from $7,500 to $60,000 per work. Judges awarded enhanced damages 35% of the time and the statutory maximum 10% of the time when infringement was willful and public performance awards were omitted. Figure 7 further explores the role of caseload composition in judge and jury award outcomes by disaggregating willful infringement awards into three subject categories: fashion designs, photos and images, and music. 85 In the fashion designs and photographs categories, award outcomes clearly followed the pattern that we observed in the two previous boxplots. In fashion design cases, for instance, the median jury award for willful infringement was $83,333 per work compared to the median judge award of $24,000. Similarly, in photos and images cases, the median jury award was $82,500 per work while the median judge award was only $3,750. Moreover, in both categories, jury awards were far more variable than judge awards, 85. We limit our analysis to these categories and to awards for willful infringement because combinations of other subject categories and types of infringement did not have a sufficient number of observations for purposes of comparison. and judge awards tended to be concentrated at the lower end of the statutory damages range, especially for awards involving photos and images.

Looking just at medians, the pattern does not appear to hold for awards in the music category. The median judge award for willful infringement involving music was $30,000 per work compared to a median jury award of $9,250. The high median for judge awards, however, is likely due to the small number of judge awards available for this analysis (n=5). A look at the distributions indicates that judge awards tended to be concentrated below the $30,000 ordinary limit whereas jury awards were widely dispersed across the entire statutory damages range, up to $150,000. The middle 50% of jury awards in music cases, for instance, ranged from $9,250 to $99,830 compared to a smaller range of $7,500 to $30,000 for judge awards. Juries furthermore made three awards at or near the maximum allowed for willful infringement: $120,000 per work, $132,500 per work, and $150,000 per work. Meanwhile, the highest judge award in a music case was $40,000 per work.

In summary, juries tended to grant higher awards than judges overall, particularly when infringement was willful. But the types of cases that were before juries and judges mattered. When public performance of songs awards were omitted from our analysis, for example, the gap between the median jury award for willful infringement and the median judge award for willful infringement decreased, and judge awards became more dispersed. When we disaggregated jury awards and judge awards for willful infringement into three subject categories, we found that jury awards tended to be higher and more dispersed than judge awards. The difference between judge and jury awards was especially pronounced in photos and images cases.

As we discussed earlier, some courts have stated that statutory damages awards should be determined according to a single-digit multiple of lost licensing fees, where such lost licensing fees are known or can be approximated. 86 To explore whether there is an association between lost licensing fees and award outcomes, we plotted statutory damages awards against the value of lost licensing fees in the subset of public performance of songs cases (n=49), television rights cases (n=5), and photos and images cases (n=10) in which lost licensing fee evidence was presented. Figure 8 first shows the relationship between awards and lost licensing fees in public performance cases. Although this plot cannot establish that there is a causal relationship, it reveals a strong, positive association between award amounts and lost licensing fees in public performance of songs cases (r = 0.67; p < 0.001). This relationship holds whether infringement was willful (r = 0.62; p < 0.001) or not (r = 0.85; p < 0.001). On average, judges awarded statutory damages amounts that were 2.4 times the amount of the lost licensing fee per work in public performance of songs cases. This average multiple is just below the multiple proposed in the case law discussed above, in which courts stated that statutory damages should reflect three to five times the lost licensing fee. A third of public performance awards, however, were between three and five times the lost licensing fee. One award was 6.4 times the lost licensing fee. 87 On average, judges awarded a slightly higher multiple of the lost licensing fee when they made a finding of willful infringement (2.6) versus when they made a finding of nonwillful infringement (2.0). The difference between these averages, however, is not statistically significant (p = 0.14).

Copyright owners provided evidence of lost licensing fees in a smaller number of television rights and photos and images cases. Figure 9 plots lost licensing fees against awards for these two subject categories. Because the number of observations here is so small, we urge caution when making inferences about these subject categories. First, the right panel of Figure 9 shows that when it comes to the nine awards involving photographs and images, there is no relationship between lost licensing fees and statutory damages awards. This is evidenced by the flat regression line, indicating that award amounts did not increase as the value of the lost licensing fees increased. Second, for the television rights cases shown in the left panel of Figure 9, we observe that award amounts and lost licensing fees were highly correlated (r = 0.99; p < 0.001). To rule out the possibility that one large award was driving this relationship, we omitted it from the analysis. Even with this outlier removed, the positive association between award amounts and lost licensing fees remained strong (r = 0.82; p < 0.001). In television rights cases where lost licensing fee evidence was presented, judges awarded statutory damages amounts that were, on average, 2.3 times the amount of the lost licensing fee per work. This number is strikingly similar to the average multiple of 2.4 in public performance of songs cases but still below the multiple of three to five times recommended in some opinions. 88

Copyright plaintiffs routinely ask a court for a particular amount of statutory damages. Indeed, as Ben Depoorter has shown, many plaintiffs employ a strategy of claiming entitlement to enhanced damages for willful infringement but asking the court for an award within the ordinary, unenhanced range. 89 Of the 277 awards in our dataset, 139 have data on the amount copyright owners requested from the court. Seventy-one of these 139 awards were associated with public performance of songs cases. Figure 10 shows award amounts and amounts requested by plaintiffs for these seventy-one public performance awards. There is a strong, positive relationship between the amount sought by plaintiffs and the amount ultimately awarded by 88. Four of the five cases represented in Figure 9 involve the unlicensed showing of pay-per-view boxing matches, which closely resemble the features of the public performance of songs cases.

89. See Depoorter, supra note 21, at 438. the court in public performance cases (r = 0.89; p < 0.001). For fortyfive of the seventy-one awards (63%), the award that was granted was the exact amount that the copyright owner had requested. These awards appear just above the regression line in Figure 10 (some of the forty-five points are obscured, however, because multiple awards overlap on the same values). The amount awarded by the court never exceeded the amount requested by the plaintiffs. Judges awarded the amount sought by the plaintiffs 70% of the time when there was a finding of willful infringement and 54% of the time when infringement was nonwillful.

Outside of the public performance of songs category, sixty-eight awards were associated with an award request. Again, there is a positive relationship between the amount requested and the amount awarded (r = 0.59; p < 0.001). Thirty-six of these sixty-eight awards (53%) were equal to the amount requested by the plaintiffs. Table 2 shows the percentage of awards in each subject category in which the amount requested by the plaintiffs was granted in full by the court. Ten of the twenty-four printed materials awards in our dataset, for instance, involved an award request. The court granted the exact amount requested by the plaintiffs in nine of those ten instances (90%). Moreover, eleven of thirty-one music awards involved a request from the plaintiffs, eight of which were granted in full by the court (73%). Plaintiffs' award requests were granted less often when the infringed works involved other (non-fashion) designs and photos and images. The court granted plaintiffs' award request in three of thirteen instances (23%) in the other designs category and in one of eleven instances (9%) in the photos and images category.

Fifty-four of the 277 awards in our dataset were associated with a damages request and evidence of a lost licensing fee. Forty-six of those awards involved the public performance of songs. Figure 11 plots the forty-six public performance awards for which plaintiffs made a damages request and put forth lost licensing fee evidence. 90 The vertical axis shows the ratio of the amount requested to the value of the lost licensing fee. This ratio represents the multiple of the lost licensing fee that plaintiffs were seeking. For example, a value of three 90. There are fewer than forty-six points visible in this figure because some awards with the same value overlap and because the y-axis is truncated at ten for clarity of presentation. on the vertical axis means that the plaintiff requested an award that was three times larger than the value of the lost licensing fee. The horizontal axis shows the ratio of the amount awarded by the court to the amount requested by the plaintiffs. This axis represents the share of the plaintiff's award request that was ultimately granted by the court. A score of 0.5 on the horizontal axis, for example, means that the court awarded half of what the plaintiff had requested; a score of 1.0 means that the court granted the full amount requested.

Figure 11 illustrates three interrelated patterns. The first involves the column of points on the right side of the chart. These points represent twenty-six instances in which the court granted an award that was equal to the amount requested by the plaintiff. In twenty of these twenty-six instances (77%), courts granted the amount requested when the amount requested was no more than three times the value of the lost licensing fee. The horizontal line in Figure 11 represents this threshold. Second, and relatedly, when the request was below the 3x threshold, courts typically granted the requested amount. Specifically, courts granted the full amount requested in twenty of the twenty-four instances (83%) in which plaintiffs requested an amount that was no greater than three times the lost licensing fee. Third, as the amount requested rose above the 3x threshold, courts tended to grant a smaller share of the requested amount. Courts granted the full amount requested in only six of twenty-two instances (27%) in which plaintiffs requested an amount that was greater than three times the lost licensing fee. Five of these six instances, however, involved requests that were between three and four times greater than the lost licensing fee. Typically, when the amount requested was greater than three times the lost licensing fee, judges reduced the size of the award relative to the amount requested (r = -0.64; p = 0.001). In one case, for example, the plaintiff requested an award that was seven times the value of the lost licensing fee. The court awarded the plaintiff half the amount requested. 91 In another case, the award request was eight times the value of the lost licensing fee. In that case, the judge awarded 40% of the amount that plaintiffs had requested. 92

Overall, our empirical analysis reveals ten features of statutory damages awards for copyright infringement. We will first summarize those features. We will then discuss what the data suggests about how well the Copyright Act's system of statutory damages has been working in practice, and what courts could do-without changes to the Copyright Act's text-to make the system work better.

Sixty percent of awards took the form of fourteen distinct dollar amounts. Common amounts were generally round numbers and often amounts set out in the statute, such as $200, $750, $30,000, and $150,000. The most common amount was the ordinary minimum of $750 per work, which was awarded 13.4% of the time. The next most common award amounts were $3,000 (6.5%), $30,000 (5.4%), $5,000 (5.1%), $150,000 (4.3%), $1,000 (3.6%), and $100,000 (3.6%).

Just over half of the awards in our dataset were $6,000 per work or less. Enhanced damages above $30,000 per work, however, were awarded 18% of the time.

3. Awards per work tended to be smaller when there were more works involved.

For the 79% of awards that involved twelve works or less, the median award per work tended to be smaller when the number of works infringed was larger. We were not able to identify patterns beyond twelve works due to the small number of observations.

Cases involving movies, photos and images, printed materials, and public performance of songs had the lowest median awards: between $1,214 and $4,063 per work. Cases involving fashion designs, software and video games, and artwork and illustrations resulted in the highest median awards: between $22,000 and $50,000 per work.

Awards involving artwork and illustrations were the most variable. The middle 50% of awards in this category ranged from $18,852 to $95,000 per work. Awards in cases involving software and video games and the public performance of songs were relatively concentrated. The middle 50% of awards in software and video games cases ranged from $30,000 to $40,300. The middle 50% of awards in cases involving the public performance of songs ranged from $2,000 to $6,000 per work.

The median award for willful infringement was $10,000, compared to medians of $200 and $3,000 for innocent and nonwillful infringement. The middle 50% of awards for willful infringement ranged from $3,000 to $60,000. In contrast, the middle 50% of awards for innocent infringement ranged from $200 to $698, and the middle 50% of awards for nonwillful infringement ranged from $750 to $7,400. Courts awarded enhanced damages 32% of the time when infringement was willful.

7. Jury awards tended to be higher and more variable than judge awards.

Overall, the median jury award was $20,000, and the median judge award was $3,775. Jury awards were widely dispersed across the statutory damages range. The middle 50% of jury awards ranged from $6,000 to $99,830 compared to a range of $1,250 to $10,000 for judge awards. When there was a finding of willful infringement, the median jury award was $50,970, and the median judge award was $7,500. When public performance of songs cases were omitted from the analysis, the median jury award for willful infringement remained $50,970, but the median judge award rose to $24,856. 8. Courts awarded between two and three times the lost licensing fee on average when such evidence was presented.

Lost licensing fee evidence was presented in sixty-six of the 277 awards. Award amounts were strongly associated with the value of the lost licensing fee in public performance of songs cases and television rights cases but not in photos and images cases. On average, courts made awards between two and three times the value of the lost licensing fee in public performance of songs and television rights cases in which lost licensing fee evidence was presented. 9. Plaintiffs' damages requests were often granted in full by the court, particularly in certain subject categories.

There was a strong association between the award amount requested by plaintiffs and the amount awarded by the court. Fifty-eight percent of damages requests were granted in full. For the subset of awards in which a damages request had been made, judges granted them in full 90% of the time in printed materials cases, 75% of the time in software and video games cases, 73% of the time in music cases, 67% of the time in fashion design and movies cases, and 63% of the time in public performance of songs cases. 10. Damages requests were more likely to be granted in full if they were less than or equal to three times the value of the lost licensing fee.

Forty-six public performance of songs awards in our dataset involved a damages request and evidence of a lost licensing fee. Courts tended to grant the full amount requested when the amount requested was no more than three times the value of the lost licensing fee. As the amount requested increased beyond three times the lost licensing fee, courts tended to award a fraction of the amount requested.

We will consider now what these observations, drawn from our dataset, suggest about how well the Copyright Act's system of statutory damages has been functioning in actual cases. The data reveals two principal grounds for concern.

First, the great variance of awards means that the extent of potential liability is difficult for parties to predict in advance. As noted earlier, the unpredictability of damages is, in some cases, a feature of the system that benefits copyright plaintiffs: the possibility of an outsized statutory damages award, one which substantially exceeds any provable amount of actual damages and profits, may be used as leverage to increase the settlement value of infringement claims. 93 On the surface, this pro-plaintiff tilt seems consistent with the overall purpose of statutory damages in copyright, which, "[b]y eliminating the burden to prove harm . . . enable the pursuit of meritorious infringement claims that otherwise would be out of reach for cash-strapped plaintiffs." 94 But relieving plaintiffs of the necessity of proving actual damages does not require that statutory damages be entirely unrelated to some approximation of likely actual damages, however rough. Nor does it require that statutory damages be as variable and, as a consequence, as difficult to predict as our data suggests that they are. While unpredictability may on balance benefit plaintiffs, it cannot in general be defended as a desirable feature of copyright law. As previously discussed, we condemn certain forms of copying as copyright infringement because we fear that unrestrained copying will depress incentives to create. But not all copying is, on net, socially destructive. Consequently, copyright law is concerned not just with suppressing socially harmful copying, but with protecting and indeed fostering copying that is socially productive. 95

The second principal concern we see in the data is the marked disparity between awards made by trial judges versus juries. Juries, in general, make statutory damages awards that are both larger, in absolute terms, and more variable across cases. It is not entirely clear why jury awards tend to be higher and more variable than judge awards. On the one hand, it may be a matter of case selectionplaintiffs choose to bring certain types of cases to juries, leading to greater variability and higher median awards. For example, plaintiffs' lawyers may recognize that juries often award high damages per work to sympathetic plaintiffs (or against unsympathetic defendants) in certain types of cases. They therefore take those cases to the jury. The higher, more variable awards in many types of cases decided by juries, however, may simply reflect a fundamental difference between jury and judge. For instance, judges are more likely to have faced copyright cases before and have regular experience determining damages, whereas jurors are typically encountering these issues for the first time. There could also be a recursive relationship: plaintiffs choose to 94. See Depoorter, supra note 21, at 403; see also In re Braun, 327 B.R. 447, 451-52 (Bankr. N.D. Cal. 2005) ("Statutory damages for copyright infringement are similar to unproven damages for violation of privacy in that actual damages resulting from such a wrong are difficult to prove, and legislatures have created a statutory remedy for this reason."); Berg, supra note 63, at 273-74 (asserting that the purpose of statutory damages is to compensate copyright owners for infringements even when it is difficult to measure actual damages).

95. See discussion and accompanying text supra notes 24-29.

bring certain cases to juries because juries give higher, more variable awards, and vice versa. Put differently, it may be that having a copyright case heard by a jury causes statutory damages awards to be substantially higher or more variable. Or it may be that lawyers perceive-accurately, as it turns out-that juries tend to make higher awards in certain cases, such as where defendants are especially culpable, and that they select those cases to be heard by juries because those are the cases for which they believe the extra expense of a jury trial is worth incurring. 96 Whether there is a causal relationship or not, the disparity between judge and jury awards is worth further investigation because it may reveal why some types of statutory damages awards tend to be less predictable than others.

A system of statutory damages-even one which, like the Copyright Act, provides for a wide range of possible damages-need not produce awards that are as variable and, as a consequence, as difficult to predict, as our data shows that statutory damages awards under the Copyright Act are. And although specification of standards for determining appropriate statutory damages awards could be written into the statute itself, they need not be. On many central issues in copyright law, the Copyright Act sets basic standards but leaves the courts substantial room to engage in common law development.

An example is the fair use doctrine set out in § 107. As the Supreme Court said in its recent decision in Google v. Oracle, "The statutory provision that embodies the doctrine indicates, rather than dictates, how courts should apply it." 97 The same is true of § 504(c) of the Copyright Act, which just as clearly indicates rather than dictates how courts should make statutory damages awards. The provision establishes ranges for permissible awards according to culpability. But beyond that, Congress has left the courts (and juries) free to make awards "as the court considers just." 98 Nothing in the Copyright Act suggests that Congress requires that trial judges or juries exercise this discretion without guidance, or that courts of appeals review an award without standards. And, importantly, the Copyright Act's legislative history explains that "there is nothing in section 504 to prevent a court from taking account of evidence concerning actual damages and profits in making an award of statutory damages within the range set out in subsection (c)." 99 The circuit courts have articulated standards, but the problem is that they are not particularly helpful in constraining variance, as is evident from the data. That problem can be addressed, but probably not through the further elaboration of multi-factor balancing tests. 100 Our data suggests that the best way to constrain variance is to gather evidence about the actual harm that the plaintiff suffered and base statutory damages on that approximation of actual damages.

The most important, and likely the most attainable, first step in reducing the variability of statutory damages awards is to try to make all copyright infringement cases look more like the subset of cases in which evidence of actual damages was made available to the judge. Judges can do this by using their existing authority to structure incentives in litigation to encourage the parties to produce evidence ). We agree with the task force's conclusion that "statutory damages should bear some relationship to the amount of actual harm suffered by the plaintiff and any financial benefits accruing to the defendant, in circumstances where these amounts are calculable." See id. at 88. Because the proposed amendment is drawn from existing jury instructions, we believe that courts currently have the power they need to anchor statutory damages awards in actual damages under existing law. See id. at 87 n.524. The task force also suggested eliminating the provision in § § 401(d) and 402(d) precluding reduced awards for innocent infringement in cases in which there is a copyright notice on the infringed work. Id. at 97. As Table 1 shows, less than 5% of the awards in our dataset involved innocent infringement, and we are skeptical that this proposal would make much difference given the low standard for willful infringement under existing law. See supra note 53 and accompanying text. Finally, the task force recommended amending § 504 so that courts do not have to award statutory damages "per work" when online services are found liable for nonwillful secondary liability. COMMERCE WHITE PAPER, supra, at 100. As noted earlier, only thirteen of the 277 awards in our dataset involved more than 100 works, and four involved more than 1,000 works. See supra note 76 and accompanying text. But this change could make a difference in atypical cases involving secondary liability, such as a recent case in which over 10,000 works were infringed. See Sony Music Ent. V. Cox Commc'ns, Inc., 464 F. Supp. 3d 795, 807-08 (E.D. Va. 2020). from which actual damages and profits can be approximated. 101 Even in jury trials and operating within the limitations imposed by the Seventh Amendment, judges may be able to reduce variability and uncertainty by encouraging the production of that evidence and then guiding juries regarding how best to use it in setting statutory damages awards.

In 66 of the 277 awards in our dataset, copyright owners offered evidence of actual damages. And statutory damages awards became markedly more predictable in many, albeit not all, of these cases. On average, courts made awards between two and three times the value of the lost licensing fee in public performance of songs and television rights cases in which lost licensing fee evidence was presented.

Of the two categories of cases in which evidence about actual damages markedly reduced variance, we have a somewhat better understanding of the underlying dynamics in the public performance of songs category. These cases generally involve performing rights organizations, which represent owners of copyrights in musical compositions (typically, songwriters and their music publishing companies). Rather than dealing with the individual artists and publishers, entities-such as television and radio stations and bars, 101. As noted earlier, § 504(b) allows defendants to recover both actual damages and "any profits of the infringer that are attributable to the infringement and are not taken into account in computing the actual damages." See supra notes 14-15. Section 504(b) adds that after the plaintiff introduces evidence of the infringer's gross revenues related to the infringement, the defendant then bears the burden of proving the expenses that should be deducted from gross revenues to calculate profits, as well as the elements of profit that are not attributable to infringement). The legislative history makes clear that actual damages and profits may be identical-that is, the defendant's profits may be "nothing more than a measure of the damages suffered by the copyright owner," in which case the copyright owner cannot reap a windfall by collecting a duplicative award. S. REP. NO. 94-473, at 143 (1976). But if there are profits that are not duplicative of actual damages, the plaintiff can recover both actual damages (to compensate the copyright owner) and the profits attributable to the infringement ("to prevent the infringer from unfairly benefiting from his wrongful act"). Id. In some cases, the defendant's profits can greatly exceed the defendant's actual damages, which could provide a powerful incentive for alleged infringers to settle. See supra note 21.

Under the approach we advocate here, judges determining statutory damages (or guiding juries in doing so) could encourage the parties to provide information allowing a rough approximation of both actual damages and, where relevant, the profits attributable to the infringement. Such information would help ensure that statutory damages awards under § 504(c) bear a relationship to actual damages/profits awards under § 504(b). This approach is consistent with current case law, which allows courts to consider the revenue lost by the copyright holder and the profits earned by the infringer. See supra note 42 and accompanying text. nightclubs, and restaurants-wanting to use musical works can purchase a license from the performing rights organization. The license allows the entity to publicly perform all the works in the performing rights organization's catalog for a certain period of time. To ensure compliance, the performing rights organization will then hire investigators to visit establishments that have not purchased a license. If the investigator discovers an establishment using copyrighted works without a license, the performing rights organization generally gives the entity a chance to purchase a license covering the period in question. If the entity does not comply, however, the performing rights organization will arrange a copyright infringement suit.

The performing rights organization plaintiffs in our dataset almost never chose to take public performance of songs cases to a jury. These organizations rely primarily on licensing fees, not litigation, to make money. But given their system of enforcing compliance by hiring investigators, they had very strong cases when they did bring actions in court. The strength of the evidence against them leads most defendants to purchase a license or reach a settlement. For infringers that do not settle, however, the case law indicating that courts should provide a multiple of lost licensing fees means that performing rights organization plaintiffs have less incentive to incur the time and expense of presenting a case to a jury. After all, the court itself will provide a sizeable punitive award of two or more times the lost licensing fees. And plaintiffs in these cases have developed a technique for encouraging courts to accept their evidence of lost fees and base the statutory damages award on a multiple of that amount. As Ben Depoorter pointed out, plaintiffs will assert willful infringement up front but will only seek ordinary damages during the damages phase. 102 Because these plaintiffs appear to be acting reasonably in cutting the defendant a break, the court is likely to go along. For all these reasons, it is no surprise that we saw strong correlations between the amount of statutory damages awarded per work and lost licensing fees per work (as well as a relationship between the amount of statutory damages awarded per work and the amount sought by the plaintiff). The handling of public performance of songs cases suggests an approach that could generalize to other case types. And courts can play an important role, by encouraging not just the copyright owner but defendants as well to offer evidence from 102. Depoorter, supra note 21, at 438. which a rough approximation of actual damages and profits could be made.

Because it can be difficult to prove actual damages, statutory damages awards need not meticulously match actual damages and profits. But for nonwillful infringement, statutory damages awards that roughly approximate actual damages and profits generally make sense. That is the view that the American Law Institute has adopted in the "Tentative Draft" of its Restatement of Copyright:

Because plaintiffs may seek statutory damages precisely because of the difficulties in proving actual damages or an infringer's profits, . . . statutory damages awards need not correspond to actual damages and profits with any precision. But in instances of non-willful infringement, a calculation of statutory damages that roughly tracks actual damages and profits, though not required by the statute, is consistent with the goals of the Copyright Act's remedies provisions and is therefore appropriate in the ordinary case. An award of statutory damages based on the plaintiff's actual damages provides compensation; an award of statutory damages that is also based on the defendant's profits that do not overlap with the plaintiff's actual damages will serve a deterrent function, as will the prospect that the defendant will be obliged to pay the plaintiff's attorney's fees in appropriate cases. 103 Further, the Restatement takes the position that statutory damages awards in willful infringement cases should be based on a reasonable multiple of actual damages and profits:

With respect to willful infringement, . . . this Restatement takes the position that statutory damages awards should be based on a reasonable multiple of actual damages and profits in cases in which a reasonable estimation of actual damages and profits can be made. A reasonable multiple is one which, in view of the circumstances of the case, is likely sufficient to deter future infringement by the defendant and those similarly situated, and to punish the defendant in a degree commensurate with the defendant's culpability. 104 Note that this approach to awarding statutory damages would not require plaintiffs to introduce evidence of actual damages or profits in every case. Indeed, courts have held that to recover statutory damages, plaintiffs are not required to offer such evidence. 105 Courts could, however, use their substantial control over the discovery process to push the parties to provide information necessary for approximating actual damages and profits, as the Restatement makes clear:

[E]ven if courts do not require plaintiffs to produce such evidence, it is within courts' authority to encourage both plaintiffs and defendants to introduce evidence helpful in establishing actual damages and profitshowever imprecise that evidence may be-in order to facilitate an appropriately sized award of statutory damages. Indeed, encouraging the production of such evidence is entirely consistent with an approach that favors plaintiffs if both parties fail to make a showing. That is, once a plaintiff has established a defendant's liability, the defendant's failure to come forward with evidence in its possession about the size of actual damages and the defendant's profits attributable to the infringement should not hinder the court or the jury in making an award-the risk that the award will fail to correspond to the plaintiff's actual damages should be borne, in that instance, by the defendant. See RSO Recs., Inc. v. Peri, 596 F. Supp. 849, 862 (S.D.N.Y. 1984) ("Any information which may be available on the exact amount of profits or damages is entirely in the hands of the defendants [. . . .] They have chosen to remain silent. Plaintiffs should not be penalized thereby."). 106 In light of the wide range of statutory damages that the Copyright Act makes available, any effort by courts to encourage the production of information by both parties about actual damages and profits would be helpful in producing awards that better track the goals of the Copyright Act's remedial provisions by reducing the risk of undersized or over-sized statutory damages awards in particular cases.

There is already substantial purchase in current caselaw that judges could use to act more forcefully in encouraging parties in copyright infringement cases to introduce evidence useful in approximating actual damages and profits. Courts already identify as among the factors relevant in determining the amount of statutory damages to be awarded "the expenses saved, and profits earned, by the infringer," "the revenue lost by the copyright holder," "the and punitive purposes, a plaintiff may recover statutory damages 'whether or not there is adequate evidence of the actual damages suffered by plaintiff or of the profits reaped by defendant' . . . infringer's cooperation in providing evidence concerning the value of the infringing material," and, most open-endedly, and therefore perhaps most usefully, "the conduct and attitude of the parties." 107 Courts could use this existing law to build out powerful incentives for all parties in litigation to introduce evidence.

First, courts could apprise plaintiffs seeking statutory damages that the court will base its award on an approximation of actual damages and that it will presume that actual damages are nominal unless plaintiffs introduce evidence supporting an approximation of "the revenue[s] lost by the copyright holder." 108 In a jury trial, the judge could explain to the jury the systemic concerns created by the wide range of statutory damages awards the Copyright Act makes available, inform the jury that it is free to address these concerns by setting its award according to a reasonable approximation of actual (in nonwillful infringement cases) or an appropriate multiple of actual damages (in willful infringement cases), and, further, tell the jury that it may presume that actual damages are nominal if the plaintiff fails to supply evidence from which an approximation can be made. Courts could go further: the factor allowing courts to shape awards according to "the conduct and attitude of the parties" would give district courts room to push plaintiffs quite forcefully to deliver up information about their business plans, their prior sales and licenses, their cost of doing business-in short, most any sort of information that would be helpful in approximating the plaintiff's damages and lost profits. 109 In jury trials, judges could advise juries regarding the judge's impression of the plaintiff's cooperativeness on these points, or the failure of the plaintiff to cooperate, and invite the jury to shape its statutory damages award in light of that information.

Courts could do much the same with respect to information obtainable from defendants, including, most importantly, information likely to be in the defendant's possession (revenues from infringing sales or licenses and the costs that are deductible from those revenues) that would be useful in approximating the defendant's profits. In cases where courts are dissatisfied with a defendant's cooperation in producing evidence relevant to damages, courts may draw conclusions on actual damages that are based on whatever evidence the plaintiff can offer, after drawing all reasonable inferences in the plaintiff's favor. Again, nothing in the law prevents district courts from threatening to defer to the plaintiff's estimates where a defendant fails to cooperate, or instructing the jury that it is free to do likewise, in order to incentivize defendants to produce information that will help in the estimation of actual damages.

Obtaining information from which an approximation of actual damages and profits can be made is a crucial first step in formulating a statutory damages award. Once the court or jury is provided with a baseline of the actual harm the plaintiff has suffered, and the actual benefit that has accrued to the defendant as a result of infringement, then a statutory damages award can be calculated that will serve the compensation, deterrence, and (in selected cases) punishment interests that statutory damages are meant to vindicate, but with less risk of enormous variance in awards untethered to the real-world consequences of a particular act of infringement. We see this in the public performance of songs awards. Outside of public performance cases, we noticed that judges often made awards involving round numbers on the continuum of the statutory range that reflected the gravity of the infringement (e.g., $10,000 for conduct that warrants a high award but not the full $30,000 per work) or numbers that are reflected in the statute: for example, $750 (the minimum), $3,000 (four times the minimum), $15,000 (halfway to the maximum), or $30,000 (the maximum). But in cases involving performing rights organizations, evidence of lost licensing fees provided an objective metric that allowed judges to anchor awards in something concrete. 110 Recently, there have been encouraging signs that some courts awarding statutory damages are first approximating actual damages, even outside public performance cases. On February 28, 2023, the newly created CCB issued its first final determination on the merits since entering operation. 111 Citing a series of cases from the Northern District of California, the district court that had referred the case in 110. Besides drawing on actual damages and profits from the case at hand, courts could also look to awards in comparable cases. To make this possible, an entity such as the Copyright Office could be empowered to compile data on statutory damages awarded by courts across the United States. And the judges of the CCB should consider-and compile-data across small claims cases, which may also be useful to parties in federal court. Compiling accurate information on statutory damages awards in federal court and before the CCB will allow judges to award like awards in like cases on summary judgment or during a bench trial. And providing such information to juries may help juries to understand the relationship between approximated actual damages and statutory damages awards, even if the jury is under no obligation to treat such information as dispositive.

111. Final Determination, Oppenheimer v. Prutton, No. 22-CCB-0045 (Copyright Claims Board, Feb. 28, 2023). question to the CCB, the CCB explained that "the general standard" used by courts "is to establish a relationship between statutory damages and actual damages. 112 The CCB also observed that courts have awarded the minimum amount of $750 in cases in which plaintiffs refused to submit proof of their actual damages. 113 In this case, the photographer claimant failed to provide evidence regarding the licensing of his work. 114 In light of "the slim record regarding damages," as well the example of comparable cases involving the photographer, the CCB awarded $1,000, a small increase from the statutory minimum of $750. 115 Encouraging the parties to provide evidence of actual damages and lost profits as a way of determining statutory damages, awarding close to the minimum when the copyright owner fails to do so, and looking at awards in comparable cases should help create a more predictable system of statutory damages going forward. And since most courts have little experience awarding statutory damages in copyright infringement cases, especially relative to the copyright experts at the CCB, the Copyright Office could compile and publish information about CCB awards that could also prove useful in traditional copyright infringement litigation in federal court. Judges could use this information to ensure that similar cases are treated similarly, and they can also use it when they instruct juries.

In this Article we have provided empirical evidence that in copyright infringement cases in which information is available that permits the approximation of actual damages, courts in many cases award statutory damages that are guided by, and are a reasonable multiple of, those approximated actual damages. Given the systemic concerns raised by the very wide possible range of statutory damages awards that the Copyright Act makes available, we have argued that courts should encourage parties to offer evidence from which actual damages can be approximated, even if only roughly. Doing so is likely