About Charles Franklin

Charles Franklin is the founder of PollsAndVotes.com, a site devoted to non-partisan analysis of the details of polling and election outcomes, mostly in the United States. In 2005 he created the site PoliticalArithmetik and in 2006 became co-developer of Pollster.com. Pollster had a good run and was bought in 2010 by HuffingtonPost. Charles is past president and an elected Fellow of the Society for Political Methodology.

Obama Approval since Benghazi hearing, IRS and AP

An update as of June 3 is added at the bottom. Still no change in approval trend, with Quinnipiac poll included. 

A lot has been written about the impact, or rather the lack of impact, of the recent Benghazi hearing, IRS and AP phone records revelations on President Obama’s job approval.  I was sitting this one out until Steve Kraske of the Kansas City Star called Thursday to talk about the issue. His piece appeared today and is nicely done.  I ran an analysis of my own in order to answer his questions and that analysis is what you see here.

The chart above shows Obama approval since January 1, 2013 using all 162 polls taken through May 23.  The red line is the estimated trend using a simple quadratic model, the smoothest trend for these data. That trend put approval at 53.5% on January 1, and at 48.9% on May 23, a decline of 4.6 points over the year so far. But as the red line shows, that decline has recently flattened and little has changed since the end of March. The red trend estimate was 48.9, as it is now, on March 26, and declined to a low of 48.5 on April 23, after which it has returned to 48.9. None of those recent changes are statistically significant.

The blue line in the chart is my standard local regression measure of approval trend, set to a moderately sensitive level. If the blue line departs from the red line, that suggests some time points where the local trend is different from the longer term smooth trend, possibly suggesting some interesting political effects. As it happens, the blue line has tracked quite closely with the red smooth trend, and especially so in the last month where the lines are all but indistinguishable.

An alternative approach compares the 16 polls taken in the 14 days before the May 8 hearings to the 17 polls taken in the 14 days after. Those data are shown in the chart below, with the means and distributions of the before and after data. The pre-hearing approval mean is 48.69 while the post hearing mean is 48.71.  Not much difference.

Given these charts, it is no surprise that a statistical model finds no evidence at all of a change in trend or mean since the Benghazi hearing on May 8 and the subsequent IRS and AP revelations. That conclusion fits well with virtually all the other analyses on this topic.

It is possible, however, that the overall trend masks some shifts among partisan groups. Democrats and Republicans are apt to react differently with shifts among independents the most important indicator. We have, thanks to Pollster.com, 43 polls since January that break out approval by partisanship. That gives us only limited information about change since the Benghazi hearing, but any such shifts may at least be a canary in the coal mine.  The statistical analysis of these data show nothing close to statistically significant, so the shifts visible in the charts below, and the estimated effects, should be treated with great skepticism. The test will come over the next month or two as data accumulate and any “real” impact will prove enduring or will wither away to nothing.

The three charts below plot the red overall trend line for each partisan group, and then overlays a relatively sensitive blue local regression line. Since the hearings on May 8, there has been a slight uptick in approval among Democrats and a slight downturn among Republicans. The statistical model estimates Dems are up 2.8 points and Reps down 3.2 points, though neither are close to statistically significant.

Among independents, the move is, as with Democrats, upward since May 8, by 2.7 points, still not statistically significant.

So the wise statistician would say “no evidence” of change among any of these groups. I would agree the evidence is far too weak to reach any conclusion of an impact yet.  But if you read tea leaves and stare into crystal balls there is the slightest suggestion that independents are not being driven away from Obama at this point.

In the rush to find instant effects of events, we look at data before there is enough evidence. But more important, the enduring political impact of events, if there are any, are not usually things we see in a week or even two but rather shifts in trends that set approval on a new trajectory, whether up or down. Rare events, such as the killing of bin Laden do produce almost immediate jumps in approval, but those events usually prove fleeting and the trend rapidly returns to its previous track. The real test of the impact of events over the past two weeks will come in six months when we can look back and see if May 8 represented a turning point after which the trend in approval shifted, either up or down, with lasting consequences. Such enduring shifts are the events that reshape presidencies, not short term bounces.

June 3 Update:

As more data are available it is undoubtedly true that there will eventually be some movement from the current trend, either up or down. Note, though, that the longer such a shift is removed from the breaking news of the three scandals, the less certain we can be that those were the reason for the shift in trend. Further hearings or new findings may yet occur but so far we are not seeing significant shifts.

A summary of the statistical models finds no statistically significant shift since May 8, either overall or for any partisan group (Dem, Rep or Ind). The Gallup daily poll, taken alone, has a not-quite significant negative coefficient, estimated at -2.1 points on approval (p=.080) but the Rasmussen daily has a positive but not significant estimate of +0.93 points (p=.36).