As many of you who read this blog know, we spend a lot of time thinking about -- and sometimes debating -- section 230 of the Communications Decency Act. We've often lamented, however, that there isn't a good compilation of all of the 230 cases (Eric Goldman, the dean of the 230 bar, has covered most, if not all, of the cases, but you have to sift through his excellent blog to find them). Well, back in 2008, I set out to create just such a compilation and after months of wrangling with my editors, the empirical study that research spawned is finally out in the current edition of the Loyola of Los Angeles Law Review.
I won't recount the entire article in this post (it's available as a PDF on SSRN), but I will highlight a few of the more interesting findings. First, here is the abstract:
In the thirteen years since its enactment, section 230 of the Communications Decency Act has become one of the most important statutes impacting online speech, as well as one of the most intensely criticized. In deceptively simple language, its provisions sweep away the common law's distinction between publisher and distributor liability, granting operators of Web sites and other interactive computer services broad protection from claims based on the speech of third parties. Section 230 is of critical importance because virtually all speech that occurs on the Internet is facilitated by private intermediaries that have a fragile commitment to the speech they facilitate.
This Article presents the first empirical study of the section 230 case law. It begins by providing a doctrinal overview of common law liability for intermediaries, both online and offline, and describes how section 230 modifies these doctrinal approaches. It then systematically analyzes the 184 decisions courts have issued since the statute's enactment. The Article also examines how courts have applied section 230, finding that judges have been haphazard in their approach to its application.
The Article closes by discussing the study's findings and by offering some insights into how plaintiffs and defendants have fared under section 230. While section 230 has largely protected intermediaries from liability for third-party speech, it has not been the free pass many of its proponents claim and its critics lament. More than a third of the claims at issue in the cases survived a section 230 defense. Even in cases where the court dismissed the claims, intermediaries bore liability in the form of litigation costs, and it took courts, on average, nearly a year to issue decisions addressing an intermediary's defense under section 230.
From February 8, 1996 (the effective date of section 230) through the conclusion of my study on September 30, 2009, state and federal courts produced a total of 184 decisions from 140 cases in which a party or the court interposed section 230 as a defense to liability for online content or acts. This is an average of 13.5 decisions per year during the thirteen-year period, with an average of 8.1 decisions each year (59.8%) holding that section 230 preempted at least one claim in the cases studied.
When I say that in 59.8% of the decisions a court dismissed at least one of the claims pursuant to section 230, this does not mean that in 40.2% of the decisions the court found for the plaintiff. The 59.8% figure only reflects decisions where the court dismissed a claim on the basis of section 230. If the court dismissed a claim on statute of limitations grounds or on the merits, I didn't count it as a section 230 win (I explain this approach in some detail in sections III and IV of the paper). When dismissals on grounds other than section 230 are included in the calculations, the results for defendants improve substantially, with defendants winning dismissal in 76% of the cases studied.
One of the best ways to assess a statute's impact on an area of law is to systematically compare the case law pre- and post-enactment. Unfortunately, this approach will not work for section 230, as courts issued only two reported decisions addressing an Internet intermediary's liability for speech-based harms prior to section 230's enactment. As a rough proxy, however, I sought to identify factors in the decisions that were germane to the question of liability under the common law, and then applied those factors in a "what if section 230 did not exist" thought experiment in order to gain insight into how section 230 has changed the liability landscape for intermediaries.
What I found was surprising. Many of the intermediaries that invoked section 230 likely would not have faced liability under the common law because they lacked knowledge of and editorial control over the third-party content at issuein the cases. Granted, I am undoubtedly overstating this because judges in the cases I analyzed did not focus on the two key determinates of liability under the common law -- knowledge and editorial control -- when addressing the application of section 230. Application of section 230 does not typically turn on those factors, so the calculations likely underreport the true number of defendants that would face liability under the common law. Nevertheless, my point in mentioning this is to note that many of the intermediaries who interposed section 230 as a defense did nothing more than provide an open platform for others to engage in speech.
Given this prediction, I note in the article that one might question whether section 230 is necessary. But this highlights one of section 230's most important functions for intermediaries: it seeks to give them the legal certainty, or in First Amendment terms, "breathing space," to facilitate the distribution of third-party speech that may contain injurious or illegal content. Without this increased certainty, risk-averse intermediaries would be less willing to facilitate the speech of others and the public would be burdened by their censorship.
Yet the data show that plaintiffs continue to file claims against defendants who are clearly protected by section 230 and who likely would not face liability under standard tort theories. Currently, there is no mechanism in the statute to deter these filings nor is there a fee-shifting provision (unlike, say, anti-SLAPP statutes). Accordingly, even intermediaries who succeeded in getting a claim dismissed under section 230 still bore liability in the form of litigation costs that they were unable to recoup from plaintiffs. This latter point is particularly concerning because the research showed that it took courts nearly a year, on average, to issue a decision addressing an intermediary's defense under section 230.
Now that I've whetted your interest in the article, go read the whole thing here. Over the next few weeks I'll be posting some of the figures and tables on CMLP's Section 230 Empirical Study Page, as well as the coding form and some of data. I plan to make the entire data set publicly available, searchable, and updatable at the 15 year retrospective/anniversary party for section 230 on March 4, 2011 at Santa Clara University that Eric Goldman is organizing. More on that later....
Update: Eric Goldman put up a blog post about the article and mentioned two statistics that puzzled him. I'll try and address his questions here.
* The study found that 41.2% of the decisions "involved anonymous content." Eric is skeptical that this statistic implies that in more than 2/5 of the cases the primary tortfeasor could not be found. I think that skepticism is warranted.
What trips a lot of people up is that litigants and judges often conflate anonymity and pseudonymity. The use of identity abstractions, such as Social Security numbers and nicknames, is pervasive in society. These identifiers are typically referred to as pseudonyms. Simply because someone uses a pseudonym does not mean they are anonymous. Traceable details always exist; it is just a matter of how hard one looks. This is especially true on the Internet, where every computer that connects to the network must have a unique IP address and where servers and routers within the network routinely log communications.
Because the degree of anonymity a user can maintain varies depending on the nature of the communication, the systems used, and other circumstances, I chose an objective measure (objective from the researcher's perspective) for this calculation. For purposes of this study, I tagged a decision as involving anonymous speech if the court stated that the speech was anonymous (courts are, unfortunately, very sloppy in using this term) or the parties asserted that the source was unknown. What I probably should have written was that "41.2% of the decisions involved an assertion that the content's source was anonymous."
As Eric correctly notes, "successful anonymity is really hard." In fact, I'd say that true anonymity is exceedingly rare in the online context and that very few of the decisions in the study involved truly anonymous speech. Because of time and resource constraints, my study didn't look below the surface of that figure. That would be a great topic for someone else to take up.
* The study also reports that "in more than half of the cases (55%), the content plaintiffs sued over was no longer available as of mid-2009." Eric notes that this suggests that "plaintiffs can get content takedowns despite 230 because the service provider voluntarily helps or the author takes the content down him/herself (or, in some cases, the court ordered the takedown)." Again, Eric is right about this.
In my view, this is another data point that shows that even those intermediaries who are protected by section 230 still face pressure -- primarily litigation cost pressure -- to remove content. When we looked at the availability of the content at issue in the cases, we found that when content was no longer available this was attributable to the defendant's voluntary removal in more than a quarter (27%) of the cases. In another 18.3% of the cases, the defendant's website was defunct. In a small number of cases (4.9%), all or some of the material was no longer available because of a court order.