The plunge of Facebook and Twitter shares in the last two days shows that both companies are hostages to investors’ unrealistic perceptions of how quickly they should grow even as they purge bots and trolls. Moving to eliminate all fake and malicious accounts, as well as making new ones very hard to register, would be scary given these inflated expectations.
Facebook fell 19 percent Thursday after its quarterly earnings announcement, in part because it missed revenue expectations, but also because it showed slower user growth, which the company said it expected to continue. Increasing spending on the platform’s policing also contributed to the gloomy perceptions. At midday Friday, Twitter was down about 19 percent due to similar concerns.
Yet both Twitter and Facebook reported growth that would have been healthy for any other company. Facebook said the numbers of its daily and monthly active users were up 11 percent year-on-year in the quarter that ended in June. Twitter also reported 11 percent growth in daily active users, and although the monthly measure declined to 335 million from 336 million a year ago, daily activity is a better measure of how many people actually use the service.
The growth occurred even though the companies are focusing on removing fake accounts and making conversations on their platforms less toxic. Eliminating all fake accounts, however, could have erased much of the expansion.
Twitter’s quarterly reports show that the share of fake accounts on the platform is stable. Five percent of the monthly active users were “false or spam accounts” at the end of 2017, unchanged from the end of 2016. The actual proportion is likely higher: In a 2017 paper, Indiana University’s Onur Varol and collaborators estimated that 9 percent to 15 percent of all active Twitter accounts are bots. But even accepting the company’s estimate, the elimination of the fake accounts would have driven down Twitter’s monthly average users, or MAU, by almost 17 million, not 1 million; the 11 percent growth in daily average users, known as DAU, probably would have been undercut, too.
Facebook reported that at the end of last year, 3 percent to 4 percent of monthly MAU were “false” and another 10 percent were duplicate accounts. At the end of 2016, it said it had 1 percent “false” and 6 percent duplicate accounts. So the share of accounts that shouldn’t really be counted toward the user base appears to have increased. Facebook only reports how many fake accounts it removes in total, including thwarted attempts to register new ones; it’s impossible to estimate from these reports how effectively it is eliminating existing fakes.
Both Twitter and Facebook could have moved decisively to weed out the fakes -- for example, by asking all users to identify themselves and declare ownership of the duplicate accounts they have set up. To avoid scaring off legitimate users by requiring too much personal information, the companies could adopt some model of anonymous identification that doesn’t allow setting up multiple identities. That would have been a big help in keeping the online conversation civil and scam-free: Identified users probably would attach a greater value to not being banned for violating the platform’s policies. The costs of fighting off automated attempts to set up fake accounts would be reduced, too.
And yet, even if the social platforms’ estimates of the current share of fake accounts are correct, doing something like that would mean reporting some numbers that investors would hate even more than they hate the current ones. And what if the estimates are low? In reality, Twitter and Facebook are unable to estimate the potential effect of introducing identification. The consequences to their valuations and to employees’ compensation would be unpredictable. Advertisers might rebel, too.
The companies will be in trouble with regulators and brands if they don’t show they are trying to fix themselves. But doing a good job would panic investors. That’s why the quality of the conversation on the platforms won’t improve fundamentally.
Bret Schafer, social media analyst and spokesman for the Alliance for Securing Democracy, which runs Hamilton 68, a tool that tracks what it defines as “Russian influence operations” on Twitter, says the platform has suspended 40 to 50 of the 600 accounts the bipartisan advocacy group started tracking in September 2017. “Without doing a full manual review of each individual account, I wouldn’t be able to give you a number,” Schafer said in an email, “but the overall volume of tweets has stayed fairly consistent.”
Schafer said that, within the general picture of Russian propaganda operations, bot activity has decreased lately. “Twitter deserves credit for that, but problems persist.”
So far, Twitter and Facebook have preferred to bleed a drop at a time, working to lower investors’ expectations of continued explosive growth. As the last days show, this tactic can backfire: A little bad news can cause steep stock price drops. That makes me wonder if cleaning up the user bases decisively would make things much worse. Investors might even welcome it, especially if, as social network users themselves, they notice changes for the better.
Leonid Bershidsky is a Bloomberg Opinion columnist covering European politics and business. -- Ed.