Instead of demonizing Nate Silver for his bias (which, in hindsight, turned out to be 100% true, not just gross outcome, but 50 out of 50 right) Republicans (now “shellshocked” about their loss) should read and absorb his new book (full of pesky facts): The Signal and the Noise: Why So Many Predictions Fail – but Some Don’t.
Being a geek with ample math background I’ve been a fan of Nate Silver since first hearing about him in 2008 election cycle. Of course for the 2012 cycle he was all over the place that I read and, of course, making the prediction I wanted to hear. So it was easy to be a fan. And also understand (but didn’t actually know about) that the Repugs demonized him with all their usual insults. I saw some of his appearances on the book tour circuit but wasn’t much interested in actually buying his book until after he began his victory laps on the talk circuit and the post-election coverage indicated that the Repugs really had predicted and believed their utterly wrong predictions whereas Nate Silver, using fact-, math-, and science-based prediction was completely right and they were completely wrong (not just lying, which is what I assumed). So, falling for the greatest marketing trick in selling books, I grabbed yet another sample for my Kindle and after reading a small amount I was hooked. So Amazon and Nate get a chunk of my money.
The reason I was hooked and wanted the book (which only part way through has continued to fulfill the promise I saw in the sample) was two things: a) post-mortem gloating over the completely failure of Repug predicting (vs Nate’s completely correct predicting), and, b) that the Kindle sample of his book started with the 2008 financial meltdown, a subject where I’ve also consumed numerous books and investigation reports, plus done my own models (owing to the fact I went to the same business school all the geeks on Wall Street, who were so spectacularly wrong, went to). I trust Nate Silver to know stats and baseball scores, but understanding the 2008 crash is a bit different and when he was right on the money there I figured this was a book worth buying.
But also there has been a lot of discussion about how the Repug pundits were so spectacularly wrong. Now when people like Karl Rove and Dick Morris and Charles Krauthhammer and even George Will made “predictions” of Romney win (actually a blowout, not even close, as according to most of them) I just assume that is the usual right/Fox propaganda. This is not what they actually believe, but it is electioneering propaganda they’re paid to spout. And when Romney’s staff and especially pollsters put out their upbeat (but unlikely) statements I figured that is the same thing – they’re just lying as part of the whole campaign of lies.
But it appears that actually did believe this nonsense. And Nate Silver, unlike the Repugs, didn’t so much “believe” (i.e. faith-based knowledge) the prediction Romney would lose, he simply did the math, and as Bill Clinton said, used “arithmetic” to base his prediction about the real universe instead of living in the alternative reality that now completely controls the right-wing. Having started from waging war on science over evolution (plus believing in fact-free religion purely on faith anyway) and expanding it to climate change it was a natural evolution (pun and irony intended) that eventually the entire right-wing worldview on all issues would become fact-free, mathematically impossible, and utterly faith-based and hence their source of huge error in prediction rather than just being the cynical product of their mendacity (also in abundant supply, but in this case not the root cause of their error).
So why were they wrong, just optimistism and/or living in their alternate reality?
No. There was bias in what they did but they actually tried to do the math. But since their alternate reality no longer involves math or science they did the math wrong.
I can’t give you a specific reference (since there are many tidbits I’ve read in last few days), but here’s my explanation of the basic story of why they actually did have a pseudo-science prediction that was spectacularly wrong.
Toward the end of the campaign (and spectacularly stupid on election night itself, as witnessed by the Karl Rove flameout on Fox) Repugs more and more complained the polls (and also wicked and biased Nate Silver) were wrong. Romney had the illusive “momentum” and their “secret” polls showed an entirely different outcome (just as Rove’s calls to the obviously unbiased source, the Romney campaign, showed even Fox’s projections about Ohio was wrong). So what were these secret polls and why were they, in hindsight, the ones that were totally wrong.
Well, to answer this we have to get down in the weeds a bit more. We all know what polls are. We hear about them every day during the election cycle. But who does polls and what actually is a “poll”. Well, the who-does is simple, lots of companies (like Omaha’s Gallup). Why do they do polls, just to get a brief mention on the news? Obviously given the size of the Gallup campus here in Omaha and their employment, polling is an expensive undertaking, so how does Gallup pay for all that? Well, it’s not just whatever money they get from TV networks. Now, those numbers, we the public see, are like the free samples of books Amazon hands out. And the samples are the whet the appetite of the people who pay to buy the entire poll, i.e. the political parties and political campaigns (and maybe a few news programs). It used to be that campaigns did their own polling but that’s too expensive and so they outsource it. So Romney’s campaign bought much more detailed polling data from Gallup than we, or even Nate Silver, ever saw.
Now why do the campaigns want that data? Because they have a very important piece of analysis to do, to turn raw “popularity/opinion” data into votes! Votes, not opinions, win elections. So when the campaign buys all the detail behind the one number the public hears on TV they have all this additional information (very much like the “exit polling” numbers so heavily reported on election night).
So simply put Gallup tells Romney’s analysts something like: 92% of African-American prefer Obama and 74% of whites prefer Romney. But even that isn’t enough: we really need to know gender (since everyone knows there is a gender gap), we need to know income and education levels, we need to know exactly where these people are (rural, urban, suburban), we need to know religion, marital status and so forth. And Gallup (or most of the pollsters) provide this in their detailed for-sale data.
Now all that data has to then be pushed through a “model.” I will bet, that even among readers of this blog and/or WordPress.com in general, most people don’t really know what a “model” is. Many may have some idea, but in fact “model” involves some very specific types of math and even some science and in voting some psychology and sociology. Ultimately though a model is math, that is quantitative with inputs and outputs, which also means today “model” usually means some kind of computer program. Now anyone who has ever programmed computers know they have zero intelligence (certainly no “commonsense”); computers do only what people who program them do, they just do the detailed number crunching fast.
So in political campaigns someone needs to create that model. Nate Silver, as he explains, has his method of constructing models. His method works. And the Romney people had their method of constructing models. And their model didn’t work. The Romney people had more “data” (aka raw “facts”) than Nate Silver did, but somehow their model evaluated those facts in some different way and got the wrong answer. WHY?
Well, in the aftermath discussions it appears to be fairly simple, some bias in the assumptions they made. Nate Silver’s technique makes none of these types of assumptions but instead uses pure statistical techniques, actually implying his “model” is simpler than the Repug models. So it was the model, not the raw polling data, that misled the Repugs.
The why appears to be a really stupid thing they did. You see the voters (not the electorate at large) were quite different in 2004 vs 2008. IOW, the probability that, say, a black, upper income, suburban, non-religious male, would vote was different in 2004 vs 2008. And so only for all the slices of the electorate based on all these factors the pollsters measure. That’s what is baked into the model that then processes the raw facts of the polls.
Now let’s think about it. In 2004 we had an incumbent and 2008 we didn’t. In 2004 the Democrat was a rich, white, Massachusetts stiff, not very likable guy, IOW, Mitt Romney (oops, I meant John Kerry). And, just in case the Repubs forgot in 2008 we had an unprecedented candidate, the first African-American and constitutional law professor to boot. Yep, John Kerry and Barack Obama definitely must be two peas in a pod. But what about the other side? In 2004 we had good-old boy, dumbshit, cowboy, bring-’em-on and heckuva-job-Brownie and in 2008 we had a stiff, rich, weirdo religion, prude that used to be a Massachusetts liberal but then became a severe conservative but then became whatever you wanted when you flipped the Etch-a-Sketch. So, once again, George W. Bush and Mitt Romney, just two identical peas in a pod.
So, despite all this “macro” (big picture) data, which election did the Repugs choose to put in their model, 2004 or 2008. Naturally they picked 2004. Now I think it is fair to say that minorities were about equally fond of Kerry and Obama, but what one were they excited about getting off their butts and voting for – Obama obviously. But the Repugs assumed that 2008 was an anomaly where minorities were all excited about their first chance to vote for the black guy (also assumed that about youth) and come 2012, all those people would return to their usual apathy. And this is want they put in their model, 2004 assumptions about voter turnout rather than 2008 turnout.
Now they actually were almost right. In fact, 9 million less people voted for Obama in 2012 than 2008, for lots of reasons, but certainly the hope and change or precedent-setting themes of 2008, vs, same-old, same-old incumbent in 2012 played a big role.
But guess what Repugs? Obama’s team made the same assumptions you did BUT they did something about it. They knew they’d get less turnout in 2012 than 2008 so they supercharged their ground game to deal with the greater apathy (among their voters that the summary polls were showing). BUT, they only worried about the ground game in ten states. Now, sure, all campaigns pay more attention to Ohio than they do Nebraska (sure red) or Massachusetts (sure blue). But, in 2012, Obama’s campaign massively bet on swing states like never before. So, in fact, they overcame the apathy factor where it counted.
So by modeling the electorate as 2004 instead of 2008 and then pushing the raw polling data through that, sure enough, Romney’s “pollsters” were seeing a few percentage points higher votes for Romney than the polls the rest of us (and Nate Silver) saw on TV. And since these numbers were what they wanted to hear, they believed them.
Now the first chapter of Nate Silver’s book covers the 2008 financial crash and in particular how the ratings agencies (S&P, Moodys) were so spectacularly wrong with their stupid AAA ratings on securities that later “facts” proved to be junk. Same thing, a self-fulfilling model. You see the ratings agencies didn’t “lie”, they just created a model which “predicted” what people who paid them money wanted to hear. And being good Repugs and believing everything is about the market and incentives (i.e. how to get paid the most) truth began to == compensation. And so it was with Romney’s people, who in turn passed on data to Rove and Morris and Will et al, who in turn all told us everywhere they could get on TV Romney would win in a landslide and the echo chamber of the right-wing media’s alternative fact- and science-free universe amplified the rosy predictions. In short, the right ran their election campaign on faith, not science, and they lost, big-time.
Watching Bill Maher the other night a fight broke out over how stupid the Repugs are in their bubble and alternate fact-free world (vs fact-free arguments this isn’t true) and James Carville actually got it right, from my POV. He said he hopes the Repugs continue to spin their stupidity and thus deny it which implies therefore they won’t fix what they did wrong (since they’ll never admit they were wrong) and therefore they’ll continue to lose elections which is what he, Carville (representing Clinton-style Dems) hope will happen.
So after all this long blog post, I reverse my title.
Republicans! Please do not read Nate Silver’s book. Please continue to believe Nate Silver lied from his bias (actually he is biased, he believes in facts and math). Please continue that Romney had momentum and would have won if it hadn’t been for Sandy. Please continue to believe Rove is a political genius and continue to give him your money. Whatever you do, stay in your alternate reality and please continue to lose elections because you trust faith more than science.