Against the Grain
Meet the One Numbers-Cruncher Who Foresees Democrats Holding the Senate
He’s not as well-known as Nate Silver, but Princeton’s Sam Wang has a method, too, and he sees Democrats holding (barely) the upper chamber.
The list of pundits, political analysts, and numbers-crunchers who are predicting Republicans will win control of the Senate in November is long, including Nate Silver of Five Thirty Eight. The folks at The New York Times’ The Upshot are saying it could be a tie. But Sam Wang of Princeton stands almost alone in forecasting that the Democrats will just barely hold their Senate majority.
Wang says he thinks the Democrats have a 70 percent chance of holding control of the Senate. As of Monday afternoon, Nate Silver thinks the Republicans have a 58 percent chance of winning, and The Upshot gives Republicans a 52 percent chance now calling it a tossup.
Wang is a 47-year-old professor of neuroscience and molecular biology at Princeton who uses advanced statistical methods to study how brain circuits work. He is the author of two books on the brain and his recent work focuses on autism. Politics, he says, is just kind of a hobby. “It’s a relatively easy problem compared with the other things I do,” he told me.
He started dabbling in politics in 2004 when he devised a computer program to aggregate and analyze polling data on the presidential race and in 2008 he founded the Princeton Election Consortium, a webpage featuring his modeling and blogs.
In 2012, his statistical analysis correctly predicted the presidential vote in 49 of 50 states and all 10 competitive Senate races including Montana and North Dakota—which, he likes to point out, Nate Silver got wrong.
A bit of a rivalry has developed between the two men. Wang refers to Silver as “the king of the nerds,” and while it may take one to know one, Silver seems a little irritated by Wang encroaching on his turf. He has dissed Wang’s Senate call and in a recent appearance on WNYC said Wang uses “arbitrary assumptions,” which Wang calls “an out-and-out falsehood.”
Silver also said he “would like to place a large wager against that guy,” not even bothering to mention Wang by name—reminiscent of the “I did not have sex with that woman” sort of dismissal.
Wang’s response: he was more accurate than Silver in 2012 and wants to “let the math do the talking.” If Silver turns out to be wrong about this election Wang told me Silver “can eat a bug.”
As Brian Lehrer of WNYC said on his show Friday—“Math Fight!!!”
What Wang does differently than Silver and other prognosticators is base his analysis only on the polling data for each Senate race. Everyone else uses other factors including their own judgment to try to enhance the accuracy of their predictions including things like Barack Obama’s popularity in different states, economic indicators, Democratic and Republican favorability, and turnout predictions.
Wang’s model uses an 80-line meta-analysis program he designed that takes all of the available polls and aggregates them to get a statistical snapshot of what is happening.
Although not every poll is reliable, Wang says even though individual pollsters can have biases and make mistakes, if you add all of the polls together they are very good at predicting outcomes.
“Let the statistics do their job and get out of their way and don’t put a finger on the scale,” says Wang in describing his method.
And this year, says Wang, those polls right now are telling him that the Democrats are going to emerge from the election with a 50-seat majority, for a net loss of five. Since Vice President Joe Biden is the Senate’s presiding officer and tie-breaking vote, the Democrats only have to win 50 seats to retain control while the Republicans have to win 51.
The top seven races—North Carolina, Georgia, Louisiana, Iowa, Arkansas, Alaska, and Kansas are all within two points right now, Wang says. He predicts Republicans will win four of those: Dan Sullivan defeating Sen. Mark Begich in Alaska; Tom Cotton beating Sen. Mark Pryor in Arkansas; Sen. Mary Landrieu losing to Bill Cassidy in Louisiana, and David Perdue beating Michelle Nunn in the open Georgia seat.
Wang and the others are also in agreement that Democrats will hold their seats in Iowa, where Democrat Bruce Braley is running in an open seat race against Joni Ernst, and that Sen. Kay Hagan will beat Thom Tillis in North Carolina.
There is also unanimity so far among Wang and his fellow handicappers about what is happening in the Colorado, New Hampshire, Kentucky, and Michigan races. Incumbent Democratic Sen. Mark Udall in Colorado is expected to hold his seat. And despite all of the national media attention former Massachusetts Senator Scott Brown has gotten by moving to New Hampshire and winning the Republican nomination there, Democratic Sen. Jeanne Shaheen currently leads that race.
Wang and the others believe Democrat Gary Peters will win in Michigan, keeping that seat in the Democratic column, and that Republican Mitch McConnell of Kentucky, who wants to change his title from minority to majority leader, is safe for reelection.
The other two states that aren’t really in play but involve Republican gains are South Dakota and Montana.
Kansas is the race where Wang differs in his prediction from Five Thirty Eight, The Upshot and others and on which his call about control of the Senate depends. Wang was onto this one before most others. He’s been doing political analysis for The New Yorker online this year, and in late August, he wrote that he thought control of the Senate could hinge on the candidacy of Kansas Independent Greg Orman, who is running against longtime GOP Senator Pat Roberts.
That assessment was based on a Public Policy Polling report that showed Orman leading Roberts by 10 points in a head-to-head matchup. Wang pointed out that Roberts’ approval rating was lower than Barack Obama’s in Kansas and suggested if Democrat Chad Taylor dropped out of the race, which he subsequently did on Sept 3, Orman had a good chance of defeating Roberts.
The latest Kansas poll shows Orman with a one point lead over Roberts—a virtual tie—with Taylor drawing as much as 10 percent of the vote (the state’s Republican secretary of state ruled that Roberts’ name must stay on the ballot a decision Taylor has appealed to the Kansas Supreme Court).
Wang predicts an Orman win, while most other analysts all have Kansas at least leaning Republican.
Add all the above up, and you get to a 50-50 Democratic advantage. But there’s a catch: The 50 would include those seats held by independent senators Bernie Sanders of Vermont and Angus King of Maine, who currently caucus with the Democrats.King said several months ago he would decide after the election whether to switch sides. It would also include Orman, who has declined to say which party he would support if he wins but who has indicated he would join whichever side has the majority. If Republicans only manage to pick up four or five seats, King and Orman could become independent kingmakers and have the power to determine which party controls the Senate.
It should be noted that Wang has incorrectly forecast Democratic victories before. In 2004, an extremely close presidential race, the polls Wang analyzed showed that George W. Bush would win, but Wang said he second guessed the analysis and tried to factor in the undecided votes, flipping his prediction to a John Kerry victory.
He vividly remembers Shirley Tilghman, then the president of Princeton, asking for his prediction. Wang told her it would be Kerry. Wang calls the flub “a humbling mistake…Now I don’t shoot off my mouth unless there are numbers involved.”
He also revealed that he was up for tenure that year and was extremely relieved it didn’t depend on his political forecasting.
Another prediction he got wrong involved the 2012 congressional election. In September of that year, Wang said he thought Democrats could win back control of the House of Representatives. Wang explains he made the mistake because at first he didn’t realize how significant Republican structural advantages were after redistricting. “I backtracked on that pretty soon thereafter. This is better than sticking with a wrong analysis,” he says.
Despite his intellectual firepower, Wang, who uses phrases like topsy-turvy and describes himself as a “politically inclined quantitative nerd,” is modest, self-effacing, and unfailingly polite. When sending a message from his phone the signature reads “Sent by handheld. Excuse brevity please!”
The son of Taiwanese immigrants, he grew up in California and earned his PhD in neuroscience at Stanford. He is the co-author of two best-selling books—“Welcome To Your Brain: Why You Lose Your Car Keys But Never Forget How To Drive” and “Welcome To Your Child’s Brain: How The Mind Develops From Conception To College.”
For this election cycle, Wang has also taken a look at gubernatorial races and has come up with a model to explain why some Republican governors in swing states could have trouble getting reelected. Those governors who stalled or blocked implementation of the Affordable Care Act are the ones Wang thinks are in the most trouble, including Scott Walker in Wisconsin, Sam Brownback in Kansas, Tom Corbett in Pennsylvania, and Paul LePage in Maine.
Wang thinks Medicaid expansion is a “proxy variable” that is predictive of extreme conservative stands on other issues as well that have hurt those governors. He points out that Republican governors in Michigan, Ohio, and New Mexico who accepted Obamacare and Medicaid expansion look to be in pretty good shape for reelection, although Michigan is close.
“I think no statistical model is perfect,” says Jennifer Duffy who follows Senate and governor’s races for the Cook Political Report. “There are subjective factors that can’t be factored in.”
If the polls were to move 1 to 2 points in the Republican direction the outcome could certainly be different but Wang says things have changed surprisingly little since June. “Democrats are outperforming expectations.” Wang says he is comfortable standing by his predictions but, ever the statistician, acknowledges: “I have a 30 percent chance of being wrong.”