The Wonderful World of Copyright Damages in Software Cases
OK, so I’ve been talking—at length—about the Federal Circuit’s recent rejection of Google’s fair use defense for its use of the way Java’s API libraries are organized. So, you might be wondering what’s next?1Alternatively, you might be completely exhausted. But hang in there; you might find this next bit interesting.
I’m so glad you asked. Let’s assume that the Federal Circuit’s decision isn’t overturned, either by the full Federal Circuit2Normally appellate cases are heard and decided by three judges, but there are many more judges serving on that particular court of appeals. Sometimes the full court decides to re-hear a case after the three-judge panel rules—known as sitting “en banc”—and can reverse or modify the panel’s decision. or by the U.S. Supreme Court3Which only sometimes agrees to hear cases.. The short answer is that the case is sent back—“remanded” is the legal term—to the trial court for a trial just on damages. That’s right: it gets sent back to the exact same court, and the exact judge, that has twice tried to rule in Google’s favor only to be reversed.
Remander Is an Unhappy Reminder
At the third trial, what will the third jury hear? Oracle will present evidence that Google owes it at least $8.8 billion in damages. Google will argue that it owes something but a whole lot less than that.
In copyright law, there are two main types of damages: compensatory damages and statutory damages. You must choose between them. You don’t get both. Statutory damages are set at between $750 and $30,000 per work infringed—not per act of infringement. The top-end of this range can be increased to $150,000 if the infringement is “willful,” and the bottom-end can be lowered to $250 if the infringement is “innocent.”4Note that these adjustments are to the RANGE. They don’t require an award of $150,000 or $250, only that the jury can pick a number between $750 and $150,000 in the case of willful infringement, or a number between $250 and $30,000 in the case of innocent infringement. Where a large number of smaller works have been infringed, statutory damages can be terrifying and lead to absurd results. However, some form of statutory damages are necessary because “actual” damages can be impossible to prove in copyright cases.
In the Java-Android case, however, statutory damages won’t be at issue. There’s only a single work at issue, Java 2 SE, so the most in statutory damages Oracle could recover is $150,000 (assuming a finding of willfulness), which isn’t much for a company as gigantic as Google (which probably sneezes out $150,000 without thinking).
Oracle has been instead seeking compensatory damages. There are two flavors of compensatory damages: “actual” damages and the “disgorgement” of the infringer’s profits. You don’t need to choose between these two types. In theory, you can recover both flavors. However, since you can’t count the same item of damages twice, and these two flavors usually overlap, it’s pretty rare to get 100% of both actual damages and disgorgement.
In a normal software case—where the defendant has pirated the copyrighted software wholesale—these damages are pretty straightforward. Actual damages are the profits you would have made if the defendant wasn’t infringing, the idea being you would have sold or licensed more copies of the software if it wasn’t for the infringement. Disgorgement would be the profits the infringing product made for the defendant. You can see how there will be a lot of overlap. Chance are, the infringing sales displaced your sales, so your lost profits are part of the profits the infringer wrongfully made. If the infringer was a lot more successful at selling the infringing product than you were at selling your product (e.g., you are a small developer and a larger company is infringing), it’s likely your actual damages are completely covered up by the disgorgement. And if the reverse is true (e.g., you’re a big company and a smaller company is infringing), it’s likely that your lost profits cover up the infringer’s profits.5It’s not impossible to recover both flavors. For example, imagine that you sell your software for $10 per unit, for a profit of $5 per unit. The software pirate undercuts you for $2, for a profit of $1, but at that price, it’s able to sell to customers who wouldn’t have purchased from you. You establish you lost 1,000 sales, or $5000 in lost profits. The pirate sold 10,000 units, for a profit of $10,000. We must reduce the disgorgement by $1,000 to reflect the 1,000 sales that are already accounted for by the actual damages. That leaves $9,000 in disgorgement to go with with $5,000 in actual damages.
The copyright owner has the burden of proving actual damages, and it can be hard to prove lost profits (because it requires you to imagine a kind of alternate reality in which the infringement never took place6This is also the pilot for the world’s most boring science-fiction program. The way the burden of proof is laid out for disgorgement is much better for the copyright holder. All you have to do is prove the infringer’s revenues for the infringing product, then the burden would shift to the infringer to prove the costs it incurred.7A perennial problem is what to do about general costs, like overhead, that aren’t directed to the specific infringing product.
While the burden of proof for disgorgement is friendlier for copyright holders, it’s not trivial, either. The copyright holder still must prove a connection between the infringement and the profits. In the normal case of software piracy, this is simple. But the Java-Android case isn’t normal. Not even close.
Java-Android Isn’t a Normal Case, You Say? No #%@¶§!
Google didn’t pirate Java wholesale, but took only a sliver of Java. Android wasn’t directly licensed for money, so no revenues are directly attributable to it. Yet, Google clearly made money from it. And, while Oracle’s predecessor in interest, Sun Microsystems, had a licensing program for Java, it was never very successful.
Oracle’s first challenge is to identify the revenue streams attributable to Google’s infringement. This is harder than it sounds, but not impossible. Oracle bears the burden of drawing a connection between those revenue streams and the infringement. Not all of Google’s revenues are related to Android, after all. And Oracle can’t just grab Google’s overall revenues from Google’s balance sheets and say, “There, I’m done with my proof.” Indeed, not all of Google’s revenues from Android are necessarily or automatically caused by its use of the way Java APIs are organized.
All the same, Oracle can get creative here. It can try to show that money Google made from Android-compatible phones—its own phones (the Nexus) and its licensees’—are attributable to Android. It can try to show that, by creating new users of Android phones, Google drove more online searches on those phones and browsing that encountered Google advertisements, such that Google’s advertising revenues were increased. This analysis is complicated by the fact that, if Android didn’t exist, at least some of those new users would have done exactly the same thing but on iPhones. But, surely some people wouldn’t own smartphones at all but for Android.
How Many Times Have You Seen the Papers Apportion the Profits?
Oracle faces another problem: apportionment. Where infringement is only partial—i.e., only a part of the infringing work uses the copyrighted work—the plaintiff is only entitled to that portion of profits related to the infringement. An easy example might be the infringement of a single song. For digital downloads, the analysis is simple: all profits for the download of that single song are apportioned to the infringement. But what about sales of the 10-song album that contains the song? Surely, the plaintiff isn’t entitled to all profits of all sales of that album. Do we just divide those profits by 10? That would be rough justice, but we could be more nuanced. What if the song were far and away the most popular song on the album? We could argue the song helped drive sales of the album, and thus more than one-tenth of the profits should be apportioned to the infringement. This would especially be the case if the song were released before the album was released and thus created anticipation for the album.
Let’s complicate the hypothetical further. What if only the “hook” of the underlying song was taken by the infringing song (and the rest of the infringing song was non-infringing)? Wouldn’t we have to apportion even more, since only a portion of the infringing song uses the copyrighted song? You could try to argue that, without the hook, you wouldn’t have an infringing song at all. But that’s illogical because it implies that songs have no value beyond their hooks, and things like lyrics, bridges, guitar solos, etc. have no value.
Google bears the burden of proving apportionment, but that won’t be too hard. Google didn’t take the entirety of Java 2 SE. It didn’t even take the APIs. It took only the way those APIs were organized, and even then, it took only a portion of that organization. Oracle could try to argue that, without the organization, there wouldn’t be Android. But that’s factually just untrue. Had Google used a different organization, it still would have had Android. Google took the organization because, when Android was new, it was worried Java developers wouldn’t develop for Android if they had to learn a new organization.
It seems to me the real value of the organization is how much more money Google made because developers found Android’s API libraries to be organized in a familiar way. This could be a lot! What if Android would’ve crashed at takeoff because Java developers just couldn’t get into Android’s different API organization? Proving this might be tricky, but that’s why big firms get paid the big money.
Previously in the lawsuit, Oracle’s expert tried to argue that no apportionment was appropriate8Based on a theory of “co-mingling.” I won’t get into it, except to say it’s way off-base., but the court wasn’t having it. Perhaps on remand, Oracle will get another bite at this particular apple.
Turning to actual damages, Oracle’s big problem is that, despite presenting evidence of pricing pressure supposedly caused by Android9I’ve already expressed my skepticism of this evidence., its expert didn’t do a very good job of calculating Oracle’s lost profits. All he did was take Sun’s assertion that its licensing revenues would increase by such-and-such amount every year and assume that any failure to meet this forecast must have been caused by Android. The court forbade the expert from testifying about these lost profits.
Oracle’s theories of damages might sound kind of fuzzy to you, but fuzzy is good enough! While the fact that you were harmed must be proved like any other element of a claim10Normally, “preponderance of the evidence,” which means that it’s more likely than not to be true., the amount of damages just has to be non-speculative. The idea is that, if you’ve been harmed, you’re entitled to something, and the fact that the exact amount of damages is hard to prove with certainty shouldn’t allow a bad actor to avoid the consequences.11An odd wrinkle is that, while damage is Oracle will have enough money to pay enough experts to get several theories of damages over this low threshold.
Damages are often a something of an afterthought, even in a big case like this with big firms, especially by defendants. Defendants don’t like to consider the possibility of losing. Plaintiffs don’t like being realistic (and given the size of jury awards, why should they be?). The way burdens shift in copyright law tends to encourage these tendencies. Oracle’s refusal to even consider apportionment is a case in point. It’s Google’s burden, after all, and apportionment looks like a concession. The problem is that, without its own viable theory of apportionment, Oracle cedes the field to Google.
One thing to look for on remand is how much additional discovery the judge allows, and how much Oracle and Google reconsider their previous positions on damages.12The parties had done a lot of work on damages before the previous trial because that trial was supposed to be on both fair use and damages, but the judge split off the damages piece fairly close to trial.
Something to bear in mind: damages-only trials tend to go badly for the defendant. The theory is that, even when juries find liability, they don’t always feel totally convinced of liability and they unconsciously carry any uncertainties over into their damages deliberations. After all, liability only requires that the jury feel it’s 51% likely. But in a damages-only trial, the jury is told to assume liability.
Should I make yet another prediction that will turn out to be wrong? Fine. I predict Oracle won’t get anywhere near the $8.8 billion it seeks.13An amount that will have increased by time of remand, since Google will have earned even more money in the meantime. And it shouldn’t, if apportionment is to mean anything.
Thanks for reading!
Footnotes [ + ]
|1.||↑||Alternatively, you might be completely exhausted. But hang in there; you might find this next bit interesting.|
|2.||↑||Normally appellate cases are heard and decided by three judges, but there are many more judges serving on that particular court of appeals. Sometimes the full court decides to re-hear a case after the three-judge panel rules—known as sitting “en banc”—and can reverse or modify the panel’s decision.|
|3.||↑||Which only sometimes agrees to hear cases.|
|4.||↑||Note that these adjustments are to the RANGE. They don’t require an award of $150,000 or $250, only that the jury can pick a number between $750 and $150,000 in the case of willful infringement, or a number between $250 and $30,000 in the case of innocent infringement.|
|5.||↑||It’s not impossible to recover both flavors. For example, imagine that you sell your software for $10 per unit, for a profit of $5 per unit. The software pirate undercuts you for $2, for a profit of $1, but at that price, it’s able to sell to customers who wouldn’t have purchased from you. You establish you lost 1,000 sales, or $5000 in lost profits. The pirate sold 10,000 units, for a profit of $10,000. We must reduce the disgorgement by $1,000 to reflect the 1,000 sales that are already accounted for by the actual damages. That leaves $9,000 in disgorgement to go with with $5,000 in actual damages.|
|6.||↑||This is also the pilot for the world’s most boring science-fiction program.|
|7.||↑||A perennial problem is what to do about general costs, like overhead, that aren’t directed to the specific infringing product.|
|8.||↑||Based on a theory of “co-mingling.” I won’t get into it, except to say it’s way off-base.|
|9.||↑||I’ve already expressed my skepticism of this evidence.|
|10.||↑||Normally, “preponderance of the evidence,” which means that it’s more likely than not to be true.|
|11.||↑||An odd wrinkle is that, while damage is Oracle will have enough money to pay enough experts to get several theories of damages over this low threshold.|
|12.||↑||The parties had done a lot of work on damages before the previous trial because that trial was supposed to be on both fair use and damages, but the judge split off the damages piece fairly close to trial.|
|13.||↑||An amount that will have increased by time of remand, since Google will have earned even more money in the meantime.|