« Forecasting Obama's chances in 2012 | Main | Twitter roundup »

November 24, 2010


In addition to comparing Republicans to hated figures, Franks uses a common trick when he accuses Reps of not caring about unemployment. I think his syllogism goes like this:

1. The Fed's purchase of massive amounts of US debt will improve the economy thus reducing unemployment.

2. Reps oppose the Fed's action.

3. Therefore Reps don't care about unemployment.

Of course, the reality is that Reps disagree with #1. We think that the Fed's action will make the economy worse, not better.

If Franks were to acknowledge the Reps' actual position, that would lead to an honest debate. He has no need to go that route, because many liberals fall for this trick and sincerely believe that conservatives don't care. In part, that's because action tends to demonstrate caring, even when that action may be counter-productive.

I will respond here to Brendan's addendum to the prior post regarding Jay Cost's post. I agree with Brendan that there's a real relationship between GDP growth rate in the 2nd quarter of the election year and the incumbant's party % of Presidential vote. The question is how good a predictor that relationship provides. I'd say not good at all. In addition to the Palin problem that Brendan points out, here are some other sources of uncertainty:

1. When the fitted line is used to prediot a single point, the uncertainty for that point prediction is different from and larger than the uncertainty of the line (which is measured by r^2.) This is a technical way of repeating Rob's point, that while the input of 2.5% matches the close election of 2000, slightly different inputs would match elections that were landslides both ways.

2. Using a predicted 2012 GDP Growth Rate rather than an actual figure adds more uncertainty.

3. The model ought to look at (Incumbant Share of Vote - Other Major Share of Vote). Looking at Incumbant Party Share of Vote means that 3rd parties mess it up. E.g. Bill Clinton's 47% of vote in 1992 was a victory.

4. The relationship between GDP Growth and Incumbant Share of Vote is likely to be changing gradually over time in some unknown way. That's why Brendan's graph doesn't go all the way back as far as data is available.

Yes, there is a lot of uncertainty around that prediction. I'm certainly not claiming that Q2 GDP growth is the best or most appropriate predictor of vote totals. (I tend to prefer the Bread and Peace model of Douglas Hibbs.) The Q2 GDP plot was just an especially simple and convenient way to illustrate that economic forecasts suggest Obama will face a very competitive re-election campaign. I'm not claiming to be making a precise vote forecast of my own.

PS David, the graph uses the percentage of the two-party vote received by the *incumbent party* as the dependent variable, which excludes third parties. Look, for instance, at 1992, which isn't Clinton. It's actually showing that George H.W. Bush received approximately 47% of the two-party vote in that election.

It's perhaps noteworthy that Jonathan Chait, who Brendan sometimes quotes, doesn't share Brendan's aversion to comparing Republicans to hated figures. On the contrary, Chait praises the attack as "the most obviously ripe political line of attack you can imagine."

The comments to this entry are closed.