August 31, 2015
Catching Up With Big Data
Posted on Jul 11, 2014
This is a column about "Big Data" and a new way to predict the results of presidential elections for the next 20 years.
We begin with the rise of Nate Silver, the former baseball statistician retained by The New York Times, and his blog, "FiveThirtyEight," which accurately predicted the results of the 2008 vote in 49 of the 50 states. He left the Times last summer to go to ESPN, a move that left the Times angry and determined. The newspaper, with a large and talented editorial staff, is famous for "swarming" big stories, that is, by overwhelming lesser media by assigning dozens of reporters to those stories.
Swarming is part of the Times DNA and when Silver left, the newspaper of record created a new department, "The Upshot," led by a very good economics columnist named David Leonhardt and staffed by 14 reporters, statisticians and editors. The department was designed to produce, in the Times’ words: "news analysis, data visualizations, commentary and historical context."
Whatever that meant, other than a lot of graphs and "visuals," last Tuesday, with some help from outside, Leonhardt and his team, along with political scientists Andrew Gelman of Columbia and Yair Ghitza, produced something that moderated my indifference to most numbers-based political science.
Square, Site wide
The Times headline on Page 3 of the print edition was: "Why Teenagers Today May Grow Up Conservative."
Well, God forbid. The analysis, however, has a ring of possible truth. The premise, in the simplest terms, was that how white Americans vote can be predicted by who was president when they were between the ages of 14 and 24. The numbers do not work when studying black voters because they tend to vote for Democratic candidates no matter what happened when they were growing up. But among white voters, the statistics indicate that events at the age of 18 are three times as likely to predict Americans’ behavior than events at age 40. After 40, it seems, very few people change their minds about politics.
The data also does not work with me as a political subject. People like me, born just before World War II, began their political lives voting 55 percent Republican and are ending up voting 54 percent Republican. The reason: We came of age when Dwight Eisenhower was president. The events of those years made most of my cohort Republicans, as did the events of the presidency of Ronald Reagan for a later political generation. I was an outlier. I adored Ike as a kid, but was pretty far left by the time Reagan was elected.
But forget me. Crunching the numbers, Leonhardt, who was educated as a mathematician, concludes that no matter how liberal today’s teenagers are on social issues, they may end up political conservatives. Why? Because, he says:
"Among today’s teenagers, Democrats do start with some big advantages. For one thing, the next generation of voters is an ethnically diverse group: About 45 percent of American citizens in their teenage years are either Latino or a member of a racial minority, compared with only 29 percent of citizens 20 and older….
"With that advantage, however comes a funny kind of problem. The Democrats are the majority party when the country is in a bit of a funk."
In other words, the numbers are telling him that if the Obama years are seen as failures, today’s teenagers will grow up as conservatives—as did the Eisenhower teenagers. Looking back, that same Big Data indicates that teenagers in the Roosevelt and Kennedy years became liberals.
Then the Times analysis concludes that all this is a huge problem for Hillary Clinton, who will promote the good economics of the 1990s when her husband was president.
"That," concludes Leonhardt, "will only go so far with voters too young to have memories of the 1990s."
Go, Big Data! It’s interesting stuff, even if I happen to think it’s wrong. Forgetting the numbers, I would go with diversity and social liberalism carrying the days 20 years from now.
New and Improved Comments