Modeling Predictions and Pomposity of People Such as Nate Silver, Sam Wang and others.

The proliferation of prediction “models” is another indication of the absurdity of the 24 hour “news” cycle. There are numerous sites making predictions about the senate races and control of that body, but the question is: what is the worth of these predictions? In my view, they are absolutely worthless and a frivolous distraction. What, one might ask, is the basis for such a view?

1. The prediction “models”  are based on poll numbers and the assumption of people such as Nate Silver that the majority of anything is better than individual positions. Silver makes this assertion in his book, and it is totally absurd. To take two current examples, we would be in imminent threat of attack from IS and we are all going to get Ebola. A longer term conclusion is there is a god who answers individual and group prayers.

2. The predictions all are based on the assumption that by taking the mean of polls,one is correcting for bad polls or ones that are inaccurate. Silver once tried to determine the worth of a poll’s accuracy but do not know if this is still true. Sam Wang is a far more sophisticated predictor and he uses the median rather than the mean, and this is in keeping with social science levels of measurement that is sometimes violated.

3. For hundreds of years, individuals and groups have tried to use arithmetic, statistical or math to make predictions and most fail if they go beyond basic statistics. The absurdity of economics and the “models” were shown in the recession, near depression of recent years. There are patterns to be explored, but there are no universal mathematical patterns that the social sciences have that can allow for cardinal rather than ordinal measures. 

4. There are interesting patterns, to be sure, that war worth exploring in politics and society, but once there are attempts to present elaborate mathematical formulas, the information becomes trivial or silly. I refer you to an article in the American Political Science Review of February 2010 or 2011, on the impact of changing precincts and voter turnout. Aside from a conclusion that a change of voting precincts by 8 tenths of a mile affected voting, the other ‘astounding’ CONCLUSION was that voting was related to party affiliation. It was article such as this one that had been funded by National Science Foundation that led congress to pass a law that prohibited Political Scientists from receiving NSF grants, and Obama signed it. I would like to see a similar law that would prohibit aggregating poll information or using it to make predictions, but that is not likely to happen. It was amusing to read that people  were rushing to read Nate Silver in the 2012 presidential election to see if Obama was going to win. But mass media leads penetration of everything leads to such absurd and irrational behavior. 

I mentioned Sam Wang’s use of the median as an improvement over Silver’s, and others, use of the mean in aggregating polls. Some predictors, I have read, have dispensed with the pompous claim—by the predictor or others—that there is some kind of “magic sauce” in predictions. By simply taking the mean of polls the predictions were as good as the more pompous, and commercial, "Model" builders with their many variable, but opaque, models. These predictions seemed to work as well as the “models’ that are constructed that run thousands of simulations. Sam Wang derides, justly so, the pompous claims about the magic sauce. Polls, he has noted, might do just as well in predicting the outcome of the senate races this year as the “models”. He, or others have noted as well, that Nate Silver made two predictions in recent years about senate races that were wrong, whereas Wang was correct. Despite Wang’s more rational analysis, and better than Silver’s as others have pointed out, I still do not have any confidence that he is any better than pollsters or astute observers  in predicting the 2014 senate results, or that can be done for the Senate this year by a two hour only work on prediction..

5. Much of the “new modeling” is hardly of a type unique to the present. I once read a 1627 book in the Cambridge University library  called Political Arithmetic and I am sure a search would find something even early based on the same theme, not to mention Plat’s Republic that seems to have the same idea in mind. There is also the widespread linking of presidential popularity with party losses. But there is a failure to tell the populace that since the Civl war, their have been fewer than 5 times where the presidential party in power did not lose seats even in the House of Representatives in off year elections, even when the president was extremely popular. Bill Clinton bucked the trend in 1998 which cost Newt his job. George W. Bush also had one off year election where he did not lose seats in the house. The same pattern has not been evident in the Senate. The new buzz word is also “wave” but V.O Key, a political scientist. Noted in the 1940s that sometimes an election resulted in a sweep by one party or the other. With regard to presidential popularity and loss of seats, Truman was very unpopular in 1948, but still won and so did the Democrats for congress. In 1954 and thereafter, regardless of his popularity, Ike lost seats in both the House and Senate, allowing the Democrats to take control of both houses of Congress again and control they did not give up until 40 years later in the House and 1980 in the Senate. The 1986 Senate races are also instructive in presidential popularity and loss of seats. Ronald Reagan was extremely popular president but in 86, he lost 8 seats in the Senate—a wave before the term was introduced I think and he never helped the Republicans take control of the House. 

6. What about 2014? Every one who looks at any media knows the Democrats are going to lose Senate seats in West Virginia, South Dakota and Montana, and are likely to lose in Louisiana, Arkansas, and Alaska so without going through all the gobbledegook of the pompous predictors, you could very well do your own self prediction that Democrats will lose control of the Senate. Maybe, to add a bit of complications but not much, the independent could win in Kansas, and the Democrats will lose in Iowa and/or Colorado, so you still have Republicans winning 7 seats and control. And you would be just as likely to be correct as Nate Silver and all the rest of the  pomposity people with so called “magic sauce”. 

7. The point is it is pretty much an exercise in taking standard information, looking a bit rationally at political culture and you have saved yourself the lab--and tedium—of following all the percentages of this and of that or Mr. percentage this versus percentage that. And you will have to spend about an hour typing up the results. 

8. And we hardly need to be told again about the importance of turn out by parties. Robert Lane, another political scientist, long ago noted that with a different fur out, policy choices by politicians would be entirely different, and we all know that by now.

8. So, no percentages of likelihood of this or that are needed,  my prediction: Republicans will control the Senate with one or two people to spare.