I went to highschool with Nate Silver in East Lansing. We were never friends, but we were briefly on the ELHS Debate team together, and I always liked him. When Nate rose to national prominence during the 2008 and 2012 elections, I was inclined to take a certain amount of hometown pride in the guy.
There were always signs that his political judgment left something to be desired. His commentary on Thomas Piketty was a travesty and his book The Signal and the Noise includes a fawning interview with war criminal Donald Rumsfeld. Still, none of that seemed to matter very much. Nate wasn’t a political commentator. He was a poll cruncher and a popularizer of important ideas about statistics and probability and he seemed to be doing good work in that capacity.
Nate Silver and Hillary Clinton
Unfortunately, in the specific context of the 2016 Democratic primary, Nate’s political judgment has started to get in the way of his primary mission. In the dispute about whether Hillary Clinton is (as she insists) a “progressive who knows how to get things done” or (as the Bernie or Bust crowd believes) a corporate shill and a dangerous warmonger to boot, he’s repeatedly come to her defense, arguing that “the numbers” show that she’s about as progressive as her opponent in the race for the Democratic nomination.
In July of last year, for example, Silver wrote that “[t]he policy differences between the Democrats aren’t all that profound; Clinton is pretty liberal…she and Sanders voted together 93 percent of the time in the two years they spent in the Senate together.”
In Fivethirtyeight’s live coverage of the MSNBC Democratic debate in February, he appealed to the same evidence to make the same point:
“If the policy differences between Sanders and Clinton seem relatively minor, that’s because they are. The two of them voted the same way 93 percent of the time for the two years they were in the Senate together, according to research by Derek Willis.”
He was taken enough with this factoid to repeat it later that night on his personal twitter account: “Clinton and Sanders voted together the same way 93% of the time.”
In his commentary on theresults in Kentucky and Oregon, he did grant that Clinton and Sanders weren’t ideological twins, but still insisted that their policy conclusions were more or less identical, using the only piece of evidence he ever uses to make this point: “They voted together 93% of the time when in Congress together.”
Why It Matters
Socialist Seattle City Councilwoman Kshwama Sawant argued in CounterPunch last month that instead of endorsing Hillary Clinton when she wins the Democratic nomination, Senator Sanders should either run as an independent or team up with Jill Stein on the Green Party’s ticket. I thought she made a good case, and I signed her petition. But if Nate Silver is right about the fundamental political similarity between Bernie and Hillary, this is nonsensical advice. Why split the vote between the two candidates if they’re in near-total agreement?
The first and most obvious problem with Silver’s analysis is that not all votes are equally important. The second is that, even if they were, the author of The Signal and the Noise would be committing an embarrassingly basic probabilistic fallacy in his treatment of the 93% factoid.
Breaking it Down
In the context of Senate voting records, there’s room for entire worlds in a difference of seven percent. Imagine two Senators who served together for two years in 2001 and 2002 and voted together differently on only two occasions in that entire time (as opposed to Clinton and Sanders, who voted differently on dozens of issues in the same amount of time). Now imagine that the three votes were on the Patriot Act and invasion of Iraq…i.e. the de facto legislative abolition of the Fourth Amendment and a catastrophic war that led to well over a million deaths and untold amounts of human suffering. Would you describe those differences as inconsequential? How about if they’d voted the same way on those two issues but voted differently on a dozen or two minor procedural votes? Would they be more different in the first scenario or in the second one?
To see whether the 93% figure is telling or misleading, we need to look at how many of those votes were matters on which every member of the caucus voted the same way, or even trivial issues commanding broad bipartisan agreement. In acting as if the raw uninterpreted frequency of Clinton and Sanders voting together means much of anything in itself, Silver is acting like a caricature of a “data-driven” journalist.
Even the source of Silver’s factoid, Derek Willis, has said that “the 31 times Clinton and Sanders disagreed [in their two years in the Senate together] happened to be on some of the biggest issues of the day…” Here are some of the votes on which they differed:
* The war in Iraq—Sanders voted against the invasion and Clinton voted for it
* A long series of votes to continue funding for the war which Sanders opposed and Clinton favored
* A long series of trade deals which Sanders opposed and Clinton favored
* The Patriot Act—Sanders voted against it and Clinton voted for it
* Reauthorizing the Patriot Act—Sanders voted against it and Clinton voted for it
* Guantanamo Bay—Clinton voted for a bill to force President Obama to keep the facility open by blocking funds for the transfer of detainees and Sanders voted against it
* The Bank Bailout—Clinton supported it and Sanders opposed it
Significantly, some of the biggest ticket items on this list, like (1) and (4), don’t even make it into Silver’s factoid, because the 93% only covers the two years when they were both in the Senate. Bernie Sanders voted against the Patriot Act and the invasion of Iraq while he was in the House. Hillary Clinton voted for both as a Senator.
This brings us to the second problem with Silver’s argument. There’s no more basic probabilistic fallacy than the Fallacy of Hasty Generalization, which we commit when we draw substantive conclusions from tiny and unrepresentative samples, and in using the 93% number to draw large conclusions about Sanders’ and Clinton’s policy preferences, Nate Silver is committing this fallacy in a particularly obvious and galling way.
Bernie Sanders has held elected office in one form or another since 1981, and he’s been in one or the other Houses of Congress since 1989. Hillary Clinton has been politically active either as an advisor to her husband or as a politician in her own right throughout the same period. As Doug Henwood has documented in his book My Turn, she spent her years in Arkansas waging relentless war on the state’s teachers’ unions. (This is what the campaign bios are talking about when they say she “fought for better education in Arkansas.”) She spent the 90s promoting NAFTA, harsher laws for criminal defendants, and draconian “welfare reform,” all of which Sanders bitterly opposed.
The problem isn’t just that Silvers’ sample size is absurdly small. The two years in which Clinton and Sanders overlapped in the Senate were 2007 and 2008. You don’t exactly have to be a political junkie to recall that Senator Clinton spent most of that time doing something other than hanging out at her D.C. office and showing up for votes in the Senate. Presidential candidates are constantly faced with trade-offs between missed votes and missed campaign opportunities. Every one understands this, and their colleagues make allowances for it. Under those circumstances, the votes she would have been most likely to show up to were those that were most important to the Democratic Party’s liberal base. Even so, she cast dozens of votes on the anti-Sanders side of “some of the biggest issues of the day.”
If you want to know whether to take a spike in Donald Trump’s poll numbers seriously or whether Ohio’s likely to go red or blue in the general election, ask Nate Silver. Look elsewhere if you want to know whether Hillary Clinton is a Sanders-like progressive or something entirely different.