New Study Predicts Facebook Will Be Dead in Three Years, Proves Studies Can Be Totally Stupid

Commentary about business and finance.
Jan. 23 2014 1:03 PM

The Myspace Fallacy

Researchers scored a viral hit with a study using epidemiology to predict Facebook’s demise. Too bad it’s fatally flawed.

(Continued from Page 1)

Moving on to Point C, things get dicier still. The authors’ conclusion relies on the premise that Facebook is already showing signs of decline. At first glance, the chart below—which was highlighted in several media reports—seems pretty convincing.

140123_$BOX_FacebookSearchQueries

John Cannarella, Joshua A. Spechle, Princeton University

But that’s very odd, given that the social network has consistently reported massive, ongoing growth in both its active user base and revenue. These widely accepted metrics, it seems, are of no interest to the paper’s authors. Instead, they pull their data exclusively from Google Trends, which measures the number of Google searches for a given keyword over time. In other words, the researchers’ claim that Facebook is faltering is based entirely on an apparent dip in the number of people typing “Facebook” into Google in 2013.

They justify this bizarre choice of proxy data by arguing that it’s “advantageous compared to using registration or membership data,” which can include inactive members. OK, then why not use daily or monthly active users, as industry analysts do? Seemingly unaware that Facebook is a public company, the researchers claim that user activity data on social networks is “typically proprietary and difficult to obtain.” They back up this assertion with a rather amusing citation: a study published in 2009 about—you guessed it—Myspace.

Advertisement

The Google Trends data on searches for Facebook, unfortunately, do not appear particularly reliable: There’s a huge spike in October 2012 that the researchers can’t really explain and end up simply throwing out. Yet they have no such qualms about an apparent dip in 2013. In fact, that dip is the key to their paper. If Facebook isn’t actually declining in popularity, then the Myspace model doesn’t fit.

The authors never entertain the possibility of an alternative explanation for a dip in the Google Trends data. Facebook’s numbers show that its users migrated en masse from using the site on a desktop computer to accessing it via their phones and tablets in 2013, often via the mobile app. The paper makes no mention of this trend—nor the even simpler possibility that people aren’t searching for “Facebook” as much on Google because they already know what it is and where to find it. By the authors’ logic, I guess a Google Trends search for “broadband” would suggest that high-speed Internet began to wane in popularity sometime around 2005.

20140123_Mbox_Broadband

Google Trends for "broadband"

Which brings us to Point D, in which the authors finally “prove” the very thing that they assumed from the outset: that all social networks can be expected to follow the same trajectory as Myspace. They even admit in the section describing their model that it doesn’t work unless you assume that all social networks eventually decline.

I contacted the paper’s authors to give them a chance to defend their findings. They sent a canned response noting that they’ve submitted the paper to a journal and prefer to withhold comment until it has been peer-reviewed. That’s their right, of course, but it did make me wonder why they published it online in the first place. Peer-reviewed or not, the study has already made its mark on the world, thanks to a bunch of media outlets in addition to Time that were happy to slap a headline on it and reap some Facebook traffic of their own. To wit:

It’s an old journalistic trick: Just add the words “research” or “study” to a sensational claim for instant credibility. Best of all, you’re absolved of any responsibility for verifying its truth, since everyone knows journalists aren’t qualified to dispute scientific findings. Funny thing, though: If it hasn’t been peer-reviewed, and journalists aren’t qualified to dispute it, then what’s to stop it from simply becoming accepted as truth?

I recently wrote about a fun blog called LolMyThesis, in which self-deprecating students summarize their research findings in a single sentence. Not all academic research can be effectively glossed in this way, but it’s amazing just how useful it can be for cutting through the confidence intervals and the literature reviews to get to the hollow core of a weak paper. I suspect that if the authors of this one thought about it long and hard, they’d have to admit that the LolMyThesis summary would go something like this: “If you assume that Facebook is like Myspace, bring in a fancy model adapted from epidemiology, and crunch the numbers, it turns out that Facebook is a lot like Myspace.”

Will Oremus is Slate's senior technology writer.