 |
| 2016 Electoral College Final Results |
Why Presidential Polls are inaccurate (not just this time around, but in general, plus factors applicable just this time around, too):
1) It's really hard to develop a model that will predict which registered voters (RV) and likely voters (LV) will actually vote, and then predict on top of that how they'll really vote. In a given state, "suburban, high school educated males" might vote one way in the suburbs of the biggest city or cities in a state, but maybe not so in smaller cities and towns, etc. Same goes for "Hispanic females", or college educated women, etc. To predict Hispanic females in Florida, for example, we have sizable Puerto Rican, Cuban, Mexican, and Venezuelan populations, distributed unevenly in terms of geographic distribution, and each of those four groups of Hispanic women is liable to have it's own predominant turnout and voting habits. So, take all the variables, and each state's population and employment nuances, a model that will accurately predict the overall result is very complicated. If the model's no good, the prediction may not be very good.
2) Assuming a reasonable model is developed, it's a lot of time and expense to collect a big enough sample to accurately model all the complexities laid out in #1, above. 500 voters isn't going to cut it. Financial and time constraints, even coming from big named pollsters means that corners get cut. If the sample isn't adequate, then the result may be a best guess, rather than something the pollster feels certain of.
3) I read that 98% of people who pollsters attempt to survey decline to participate. You've got to ring a lot of phones to get a sample of 5,000 or 10,000 or more. A lot! Then you have to wonder if those who do participate represent the whole of the voting public very well, or if they don't. If the participants aren't reflective of the voting population at large, oh well.
4) People lie to pollsters. "Shy Trump voters" and people who will simply get a kick out of being dishonest with a pollster are really out there. If a pollster doesn't craft their questions carefully to elicit reliable responses, fuggedaboutit.
5) Pollsters often have their own political motivations in conducting a poll. Whether it's their own bias or the desires of their client, often a poll's purpose is NOT to accurately predict the outcome as best as the pollster can, but it is to influence the electorate. I believe we have been seeing a LOT of that thus far in this election cycle. The pollsters working for big named media entities, New York Times, Fox News, ABC, CNN, and other big names like Sienna, Trafalgar, Rasmussen, etc., are not immune to this. What I've often observed is that the numbers move in the week or two preceding the actual election, this is an attempt by the big polling entity to move away from the influence polling numbers they've been releasing, toward something more accurate to protect their reputation as a pollster for posterity, and the next cycle. Watch the polls the next week and the following. Are the numbers changing? Why? Most everyone in this cycle has their mind made up, and has for a long time. Big changes tells you that they weren't being straight with you to begin with.
6) National numbers are nearly meaningless. Saying one candidate leads nationally by 4 points is damned near useless information. We have an electoral college. That's how it works. Ignore national numbers unless the margin in the day or two leading up to the election is huge, like 10+. I've got news for you, it won't be. If a pollster is telling you it is, which of 1-5, above, is the reason?
So, what does Good Ol' Matty P recommend you watch if you want to get a feel for how things are going? Pretty simple. Biden is going to win NY, CA, IL, MA, HI, and several others. Trump's not competitive in them, and you can chalk them off already. Trump is going to win a lot of smaller southern and western states. A lot of them: AL, MS, LA, KY, TN, etc. Probably the easiest thing to do is to look at the 2016 final map, which I posted, above, and think, "Which states has Trump got a realistic chance of flipping to red that Clinton won in 2016? Which 2016 Trump states does Biden have a realistic chance to flip blue?" If you take the sure bets off the map, there aren't near as many to think about.
Me? I think it will be close. Of the states that come in early, if Biden wins Florida, Trump will have a long uphill battle. Spoiler: Trump is going to win Florida by 3%, maybe more. And if Trump wins Florida, watch North Carolina, Pennsylvania, and Michigan. If Biden sweeps them, it'll be a long night for Trump. If Trump wins 2 of those 3, in about any combination, it's going to be a bad night for Biden. If Biden wins Florida AND Georgia (he won't), Trump is toast.
As the evening unfolds, Wisconsin, Minnesota, Arizona, and Nevada may be interesting. Good Ol' Matty P thinks Biden will need all of them to win to win it.
Things that would surprise me? Biden winning Ohio, Iowa, or Texas. Trump winning New Mexico, Colorado, or BOTH of Arizona and Nevada. I think he wins one of the two, but not both.
Polls that I'm watching now? Florida, but mostly because I live here. PA, MI, WI, MN, AZ, NV, NC: I think the race will be won or lost there.
Pollsters I'm paying attention to? Big Data Polls (People's Pundit on Twitter and YouTube), Trafalgar, and Rasmussen.
One last word: all of the above is meaningless if we don't exercise our rights and vote. So, no matter who you like or don't like, who you support or don't support, which way you lean, get out and vote. It's too easy not to. And remember: Republican Election Day is November 3rd, and Democrats, your polls will be open and ready for you on November 4th. Just kidding. But if you fall for that, you don't deserve to vote, anyways.
Have a great evening friends! God bless.
Nice read Matt!!
ReplyDelete