We can say, with reasonable confidence [85 percent], that a Republican will be moving into the White House in 2017.
That conclusion is based on the results of a data model we created, and is primarily the result of two factors…. First, a Republican will win because voters typically shy away from the party currently in power when an incumbent isn’t running. In fact, a successor candidate is three times less likely to win. Second, President Barack Obama’s approval ratings are too low to suggest a successor candidate will take the White House.
Remember when Barack Obama lost the 2012 election because of data models based on the high unemployment rate?
I'll stack my political instincts against Ipsos's model any day — and I'd bet big.
Their methodology seems to rely on the same sort of analysis that proves the Curse of Tippecanoe. Your instincts I believe are sound. Frankly this is why I believe, barring some disaster, that Hillary has the best chance of whooping the Republican candidate's ass. There's no doubt from what I've read that she'll lose some of the male vote. But I suspect she'll do extremely well with the female vote and that has underpinned Obama's back to back success. Bernie's been pretty clear that he ran as a Democrat because he didn't want to go independent and give the victory to the Republicans through a vote split. He has declared that he will support the winning candidate. And if anyone can bring his tribe on board with him I think Bernie can. I am beginning to wonder how good a vice president he would make. Pretty good is my first impression.
Posted by: Peter G | October 16, 2015 at 01:55 PM
Clickbait nonsense, but at least Ipsos used more than one criterion. There were models showing Bush I wouldn't win because no sitting VP had been elected president since Martin Van Buren in 1836.
Posted by: Bob | October 16, 2015 at 02:33 PM
These models have the flaw of any statistical model: They assume the future is like the past. (Econometrics was one of my Ph.D. fields, so I know at least a little about this.) If the underlying structure of what you're modeling has changed, the model is useless. The sizable Democratic demographic advantage in presidential elections is something that's emerged gradually over the last 20 years. So is the stark polarization of the electorate. There are now few voters truly up for grabs. To ignore these developments is to generate nonsense.
There is also the problem of applying international regularities directly to the U.S. Few countries--none I can think of--have the stark difference in presidential and off-year electorates that characterize the U.S. (Indeed, the concept of "off-year" has no meaning in a parliamentary system.) What does it mean to say that there is no "incumbent candidate" in a parliamentary system? That the former prime minister has stepped down? That definition essentially conflates absence of an incumbent with the already established political failure of the ruling party.
In short, a statistical toolkit is no substitute for thinking carefully about your modeling approach, nor for applying basic common sense.
Posted by: Matt | October 16, 2015 at 04:48 PM
Our company has developed a highly sophistimacated data analysis system that takes into account *ALL* facts and all possible permutations, and simulation after simulation could be saying the same thing over, and over, and over again: "You will win". Please deposit $10,000,000.00 US Dollars. Thank you, and God Bless you and your family and God Bless America!
Posted by: Marc | October 16, 2015 at 07:33 PM
I'd rather go with the people who are wagering actual money on the contest in the Iowa Electronic Market. For the past three months, they've pretty consistently put the odds of the Democratic candidate winning the popular vote at ~60%. Since the first Democratic debate, the price of Democratic stock has actually increased slightly, as seen in the graph at the link.
https://iemweb.biz.uiowa.edu/graphs/graph_PRES16_WTA.cfm
The Iowa Electronic Markets have a very good track record for calling the popular vote winner going back to 1988.
Posted by: Neon Vincent | October 18, 2015 at 11:05 AM