Doctor Cleveland's picture

    Predictable Results

    The Eastern Seaboard is getting clobbered by a combined late-season hurricane and blizzard, flooding large areas and knocking out electricity in even larger areas. As I write this, the New York City subways are flooded, there has been an explosion at a Con Edison power station, and a large part of the Rockaways is burning while firefighters, trapped by the floodwaters, are helpless to stop it.

    Once-in-a-lifetime storms like this one have always existed. But the increasing frequency of massive storms, coming much much faster than only once a lifetime, is one of the results predicted by the standard model for catastrophic global warning. That doesn't mean that Hurricane Sandy (aka the Frankenstorm) is necessarily a result of climate change, the way the shrinking polar ice obviously is. But climate change models do predict that there would be many more storms like this, and the last ten years have certainly had a hell of a lot of fierce weather events. Of course, many people in the media denounce the very idea of climate change as "junk science"

    The storm is knocking out millions of people's electricity a week before a presidential election. This is largely because we have allowed our infrastructure to age and weaken, until our bridges and roads and power grids are too creaky and brittle for an emergency. The problems with this have been predicted for many years, but those who advocated public rebuilding of our national infrastructure were denounced as spendthrifts, wasting the public's money and "holding back" the entrepreneurial economy.

    The election itself is on a knife's edge, largely because the economy is in the doldrums, despite the half-sized stimulus package President Obama passed at the beginning of his term. Some economists, most famously Paul Krugman, predicted at the time that such an undersized stimulus (less than half the size of the demand that the American economy had lost in the downturn) would fail to spur sufficient growth, make future stimulus packages politically impossible (because the first would "prove" that stimulus spending "doesn't work"), and lead the President into an uphill slog of an election campaign in a depressed economy. All of those things have happened. But for saying so in advance, Krugman was widely called unreasonable, partisan, and shrill. He continues to be called those things.

    This week, as the election remains close and different approaches to reading polls yield different indications about which candidate is more likely to win, a wide range of media voices ranging from professional conservatives at the National Review to self-described centrists like David Brooks and Politico launched on attack on the poll analyst Nate Silver of for the high crime and misdemeanor of favoring state-by-state polling data over the headline numbers from national tracking polls. What apparently provoked this pack-mentality assault was that Silver did not endorse the idea that Mitt Romeny was continuing to surge in the polls after, well, after Romney ceased gaining in the polls.  Eventually, some of Silver's attackers (some of whom had gotten very personal indeed) realized that nearly every poll-averaging outfit was producing results much like Silver's; this led some of them (such as noted statistical genius Jennifer Rubin) to denounce poll averaging itself.

    There's some grim humor in the attack on Silver, such as his critics' displays of colossal mathematical illiteracy (some of his attackers cannot distinguish between predicting that Barack Obama has 70% chance of winning the election and predicting that Obama will win the election in a 70-30 blowout), and their even more colossal displays of hypocrisy (various pundits who have been horribly mistaken over and over, without an ounce of chagrin, have pompously declared that if Silver is wrong about the upcoming election he will be permanently disgraced. Disgraced, I say! Just like Joe Scarborough was that time that he was wrong about every single thing that happened in the Bush Administration.) But there's something simply grim here, too, because the sudden outburst of rage suggests how our media actually works and how they treat the business of prediction.

    If Nate Silver's approach to handicapping the election is wrong (and of course it could be), we'll know in a week. And Silver's critics, if they only believed he was wrong, could just wait for him to be proved wrong, the way I wait out loudmouths in sports bars. But simply not predicting that Romney would win after the media storyline had changed to Romney's Irresistible Momentum was enough to provoke a wide-spread attack. To predict something that the establishment finds politically inconvenient is considered an outrage. And to base that prediction in rational analysis of data only makes it more unacceptable. When people like the pundits who attacked Silver this week talk about "facts," they don't mean facts. They mean socially-produced ideas of reality, established by influential insiders and handed down to the rest of us. Anyone who publicly contradicts that socially approved "reality" is disrupting business and inconveniencing the powerful, and that cannot be allowed. Carefully-constructed reality-based models only make the crime worse.

    The global-warming hypothesis, based on careful analysis of enormous quantities of data, is not acceptable to the powers that be, because accepting that reality would force them to change the way they arrange things, the way they conduct their business and make their political decisions. Those arrangements outweigh any facts. The reality of our crumbling infrastructure is likewise inconvenient, because it would require sensible taxation and a rational approach to public finance. Paul Krugman's model for the how large the stimulus needed to be was likewise inconvenient, because it did not fit in with existing political agendas. How dare he make predictions without taking into account what K Street considered reasonable?

    Our country has allowed the perspective of insiders and elites, a perspective that treats ongoing power games as the most fundamental reality, to crowd out actual reality. When science, mathematics, or simple common sense threatens to interfere with the results of those games, or render them moot, then science, math, and common sense are ruled out of order. But there's a price to be paid for running the world as if Beltway games were more real than physics, and that price is extremely high. The price is that our country fails to deal with easily foreseeable problems, and instead makes them vastly worse than they needed to be. After a while, it means our system runs a danger of failing completely. And that failure is all too predictable.



    I wish it weren't so, but I think you are the quintessential Cassandra here. I'm listening, though. And I think you're spot on. 

    "When people like the pundits who attacked Silver this week talk about "facts," they don't mean facts. They mean socially-produced ideas of reality, established by influential insiders and handed down to the rest of us."

    So true, Doc, and powerfully written.  I'm with Orlando, above.


    David Brooks is a pompous idiot. In April 2003 he wrote: "Now that the war in Iraq is over, we'll find out how many people around the world are capable of facing unpleasant facts...The ruling class is reasonably candid about the war's progress. The anonymous people in the corridors of power basically seem to know what they are doing..."

    Yes, that was one month after the invasion of March, 2003. Brooks could easily predict the war was 'over'. A war on the other side of the globe, one month into it, in a country and a region with a culture and a history about which he knew jackshit.

    Now with polls, for Brooks its: "If there’s one thing we know, it’s that even experts with fancy computer models are terrible at predicting human behavior."

    Bloody conflicts 'over' = predictable.  Voting = unpredictable

    Brooks is a good reason to cancel your subscription to the nations 'paper of record.'

    Remember, if you want to read the NYT online, you can read as much as you like for free by either deleting the NYT tracking cookie or using your browser's private browsing mode.  This works on mobile devices as well.  Now there's no need to subscribe!

    Thanks for the tip DF. 

    BTW great post Doc. C.

    Thanks, NCD!

    Climate change did not create Sandy, it merely allowed it to be created. It did not cause Sandy to  devastate the eastern seaboard, it allowed it too.

    There is a difference albeit a small one.

    I will be the loudmouth in the bar this morning, the guy who breaks into the conversation with a contrary position on a small point even though he agrees with almost everything you say.

    If Nate Silver's approach to handicapping the election is wrong (and of course it could be), we'll know in a week.

    No, we won't. We will not know on the basis of results whether his predictive method was right or wrong because his prediction clearly allows for a Romney win. If you have one roll of one die and must bet your house that it either will be a six or will not be a six, the only sensible action is to bet that it will not be a six even though you clearly understand that it could be. You bet based on  easily understood, obvious good odds that five winning possibilities against one losing possibility is the way to go. If you become homeless it is not that you were wrong in how you bet, it is just that the less likely possibility is the one that happened. Silver digs a lot deeper to understand the components that determine what he thinks the odds are but he still only produces odds based on incomplete evidence.

    NFL odds-makers are amazingly good at their predictions but they miss a lot of the time. They may predict that the deciding difference is that the winner will be the fast team playing on a fast field only to have a freak storm blow in a few minutes before kick off and coat the field with snow. Weather can screw a lot of things up, another thing the Right is wrong about. 

    If Obama wins we will be more assured that Silver's methodology is correct but I believe that considering the result itself as proof  that his methodology is correct is a logical fallacy, just as believing his methods are worthless would not be proven by Romney winning.

    All that said, I am betting, as well as hoping, that Silver's higher likely predicted result is the one that comes about, but if Romney wins I will not consider it proof that Silver's methods are wrong any more than I will consider it proof that Rubin knows from Shinola. Even if Romney wins I would still bet that his prediction would be the best one next time. It is the way to bet if your source of information is his or Rasputins', uh, I mean Rasmusson's.I would still bet against a six coming up again if my bed in the flop house depended on being right.

    I know your entire thesis does not rest on the one sentence I disagree with. Good blog.



    Thanks, Lulu. You're absolutely right.

    That sentence oversimplifies for brevity's sake, and it has all the problems you name.

    You think you're sooooo smart. That might all be true in math world, but hello! real people don't live in math world. Real people live in media world.

    In media world, if Obama wins, then Silver is obviously a genius. And if Romney wins, then Silver is obviously a biased liberal hack.

    In reality, we will know how accurate Silver's model is based on how closely he calls the EV total, which is tantamount to calling the map.  Of course, PEC has an open model that is essentially a straight average of state polls with no econometric terms.  They called 2004 on the nose and were off by one EV last time.

    IMHO, the threat to Nate isn't mouthy pundits who think math is for effete liberals.  It's that there might be a simpler, non-proprietary model that does just as good a job as his, if not better.  I don't think this is lost on him though.  I have recently seen him remark that you might as well simply average the state polls and then go about your day.

    I am really interested to see if this has any significant effects on punditry in the future.  If there exists a simple model for accurately predicting the outcome of Presidential elections, that could potentially have a very big impact on punditry, campaigning or even finance.  What might change if that notion penetrates the public consciousness and becomes common knowledge?  Consider, for instance, that Romney has never once during this cycle been the predicted winner of either of these models.  How might that affect a candidate's ability to raise money in the future?

    Great post!

    Predictions; accurate predictions should be rewarded.

    But there are negligent predictors and just flat out lying predictors.

    I was thinking about this while I view the coverage of Sandy and its aftermath.

    The meteorologists were right on; from the beginning in terms of the power of this storm and its interaction with other weather systems present as it hit the mainland.

    There are currently 8 million without power; the experts predicted up to ten million without power.

    And other predictions of devastation seemed to have come true.

    These people knew what they were doing.

    Governmental units; state, local and Federal governments have been cooperating with each other and basing their preparations upon the weather experts.

    This is what is supposed to happen.

    Silver was right on in 2008 and other pollsters like Rassmussen were way way off--up until the very end when they kind of 'cleaned up their act' to look plausible.

    So my bets are on Silver.


    This is an excellent blog, Dr. C.  If the press corps did even a basic job of reporting the relevant facts, the entire Romney campaign would be untenable.

    Thanks, DF.

    Latest Comments