But, one thing about statistical models of elections is that they kind of suck. There aren't a lot of observations so people often use data from the distant past to predict the future without testing to see if the data generating process is constant over the time period.
Also people use a lot of strange and potentially endogenous explanatory variables. As a rule of thumb, if you let me pick the dependent variable and also allow me to create my own made up right hand side variables, I'm going to get a pretty good in sample fit for my model.
So lets just look at the relatively recent raw data and see what we can see. This graph is from my 2008 Public Choice paper. It shows that from 1961-2004, there is a pronounced difference in economic performance over the second half of the election cycle between cycles where the incumbent party wins and when it loses.
I used Fred to update the results through 2008, and average growth over the last half of the election cycle when the incumbent party wins is 4.6%. When the incumbent party loses, it's 2.2%. That difference is significant at the 0.001 level.
For the relevant period of the Obama cycle for which Fred lists data (2011 q1 - 2012q2, 6 quarters), the average growth rate is 1.9%
In recent times, the incumbent party will, on average, lose with such a low growth rate in the second half of the cycle.
Why do I care about this? Well, I want to question the idea that the economy is helping (or at least not hurting)Obama. I want to question the assertion that campaigns don't matter because the economy is everything.
BHO is going to win in November DESPITE the economy, largely because Romney is a terrible candidate who is running an undisciplined, juvenile, brain-damaged campaign.
And that matters.
7 comments:
As much as I dislike BHO, four more years of divided government will be better than Romney running an undisciplined, juvenile, brain-damaged White House.
I loathe BHO as well, but consider Romney to be reasonable - although i'd agree, terrible campaign thus far. What part of it is "brain-damaged", however?
I'd like to see some error bars or confidence bands around the mean growth lines.
Paper also gives medians, that's about it though.
As I have argued already in http://mythdesisyphus.wordpress.com/2012/07/31/on-handicaping-the-coming-election/#more-497, it is neither the economy nor the campaign that has determined the election outcome in the case of a sitting president seeking reelection, at least since WWII.
It is rather the presence or absence of a credible challenge from the president's own side. So, yes, BHO is heavily favored to win this one, unless he manages to make history again by snatching defeat from the jaws of victory.
You write:
"But, one thing about statistical models of elections is that they kind of suck. There aren't a lot of observations so people often use data from the distant past to predict the future without testing to see if the data generating process is constant over the time period."
How does what you do here help you with that? If other models are crap than I see no reason to believe that this one is crap too.
Dear Sis: whether the President is challenged in-party seems to me to be pretty endogenous. That is something we might explain by economic conditions or scandals, not take as a given to predict outcomes.
Anon 7:29 / 9-19.
I am just looking at the raw data and comparing averages of winners v. losers. I wouldn't call it a model.I am dealing with DGP issues by not using the distant past. Sure there can always be a regime shift and that's a potential criticism of anything.
Post a Comment