« LA Times goes "he said," "she said" on iPhone "curse" | Main | Twitter roundup »

July 22, 2010


At Pollster.com, a commenter asks the somewhat facetious question what is the correlation between Obama and Carter. But it's a pretty good question. Gallup helpfully offers an interactive gadget that lets one graph presidential approval ratings for selected presidents over a selected time frame.

Here's a graph of the approval ratings for Presidents Carter, Reagan, G.W. Bush and Obama over the first roughly eighteen months of their presidencies. Just eyeballing the lines, they all look pretty similar, except that Bush experienced a big leap after 9/11 and his approval thereafter eroded from a much higher level--though the rate of erosion seems similar to the other presidents shown.

Brendan could no doubt download the raw data and calculate the degree of fit. Maybe he's already done that. Is the correlation to Reagan really stronger than the correlation to the others, and if not, what's the point of comparing Obama to Reagan rather than Carter, etc., other than to rally dispirited Democrats?

Rob, you're right that Reagan isn't the only possible comparison -- I posted earlier on how Obama's approval trajectory is similar to five of the past seven presidents, including Carter. The reason I highlighted Reagan is that his initial legislative success and subsequent economy-induced slide are similar to Obama and therefore a useful reference in helping people understand the administration's current situation.

Is the correlation to Reagan really stronger than the correlation to the others, and if not, what's the point of comparing Obama to Reagan rather than Carter, etc., other than to rally dispirited Democrats?

Brendan has already compared Obama to other presidents (here for example: http://www.brendan-nyhan.com/blog/2010/06/matt-bai-wrong-on-presidential-approval.html )

"Obama's approval trajectory (in purple) is tightly clustered with five of the last seven presidents." -Nyhan

And the reasons to compare him to Reagan include:

1.) Both started with nearly identical approval ratings.

2.) Both were elected in large part due to a repudiation of the previous administration's economic policy.

3.) Both saw serious recessions during their first term.

4.) The unemployment rate in Reagan's first 18 months went from 7.5%-9.6% (It topped out at 10.8% 6 months later)

5.) The unemployment rate in Obama's first 18 months went from 7.7%-9.5% (peaking around 10.1% so far)

In other words, they governed under very similar circumstances and their popularity have followed nearly identical trajectories.

Reagan, 30 years later is seen as a highly successful president. But he wasn't seen that way at the time. We won't know how history will view Obama for a while yet.

I'd forgotten the earlier comparison, for which I thank you. Given the substantial similarity of the approval trajectories of Obama, Carter, Reagan and G.W. Bush (allowing for the 9/11 boost), emphasis on the comparable economic circumstances of Obama and Reagan runs the risk of imposing a narrative on the data, and we may find ourselves nicked by Occam's razor.

Brendan's analysis is more methodical than the pundits he criticizes, but to what degree is it actually science? I find myself comparing Brendan's work to the statistics work my wife does in support of medical science.

Political Science draws causitive conclusions from much smaller samples then medical science would require. That's natural, because there aren't that many past Presidents to sample from. Furthermore, Political Science doesn't matter as much. If medical science gets cause and effect wrong, people may die, whereas projectiions of Presidential popularity are not life and death.

Howevever, the one thing I would like to see from Political Science is a clear statement of their hypothesis. Saying that Presidential popularity depends mostly on economic conditions is vague. How does it depend?

In other words, I'd like to see a formula something like:

Presidential Popularity =
f(unemployment, GDP growth, inflation, time in office,...)

A specific hypothesis like this could be tested against each past President as far back as one felt was relevant.

Without a specific formula, I worry that Brendan may be creaing an ad hoc narrative for each past President -- that is, he can show the importance of economic conditions for each past President's popularity, but in a different way for each past Presidents.

David, perhaps you and I should start a listserv of right-of-center commenters to coordinate our comments and collude on giving things the proper spin. We could call it the NyhaList.

It's my impression from reading Bernstein's blog that political scientists do have a formula that can spit out projected approval ratings within a few percentage points of actual approval ratings. I'm not a poli sci guy though, so I couldn't tell you what this formula is called

The comments to this entry are closed.