The US polling industry has been suffering a crisis of insight over the past decade or so; its methods have become increasingly bad at telling which way America is leaning. ... The classic pollster’s technique known as random digit dialing, in which firms robo-dial phone after phone, is failing, because an ever-dwindling number of people have landlines. ... whereas a survey in the 1970s or 1980s might have achieved a 70 percent response rate, by 2012 that number had fallen to 5.5 percent, and in 2016 it’s headed toward an infinitesimal 0.9 percent. And finally, the demographics of participants are narrowing: An elderly white woman is 21 times more likely to answer a phone poll than a young Hispanic male. So polling samples are often inherently misrepresentative. ... Today’s polling landscape appears so fraught that Gallup, long the industry leader, opted out of presidential horse-race polls this year; the reputational risk of being wrong was simply too high. Civis, on the other hand, promises a paradigm that could rescue American politics from confusion. ... Today, campaigns realize they have to look elsewhere for their intelligence, which has caused a major change in how the political industry functions. In the past, an entire campaign’s data and infrastructure would go poof after Election Day. Now Civis and similar firms are building institutional memory with permanent information storehouses that track America’s 220 million–odd voters across their adult lives, noting everything from magazine subscriptions and student loans to voting history, marital status, Facebook ID, and Twitter handle. Power and clients flow to the firms that can build and maintain the best databases of people’s behavior over time.
For the members of Congress, who in 2002 provided almost $4 billion to modernize voting technology through the Help America Vote Act, or HAVA—Congress’s response to Bush v. Gore—this probably wasn’t the result they had in mind. But voting by computer has been a technological answer in search of a problem. Those World War II-era pull-lever voting machines may not have been the most elegant of contraptions, but they were easy to use and didn’t crash. Georgia, which in 2002 set out to be an early national model for the transition to computerized voting, shows the unintended consequences. It spent $54 million in HAVA funding to buy 20,000 touchscreen voting machines from Diebold, standardizing its technology across the state. Today, the machines are past their expected life span of 10 years. (With no federal funding in sight, Georgia doesn’t expect to be able to replace those machines until 2020.) The vote tabulators are certified to run only on Windows 2000, which Microsoft stopped supporting six years ago. To support the older operating system, the state had to hire a contractor to custom-build 100 servers—which, of course, are more vulnerable to hacking because they can no longer get current security updates. ... The voting technology business, after a frenetic decade of mergers, acquisitions, and renamings, is dominated by just a few companies: Election Systems & Software, or ES&S, and Dominion Voting Systems are the largest. Neither has much in common with the giants of computing. Apple, Dell, IBM, and HP have all steered clear of the sector, which generates, according to an analysis by Harvard professor Stephen Ansolabehere, about $300 million in annual revenue. For context, Apple generates about $300 million in revenue every 12 hours.