After a flurry of new polls in the last 24 hours, a new Harris poll taken 4/20-23/07 finds approval at 28%, disapproval at 70%. With this addition, the estimated approval trend stands at 34.1%.
The Harris result is a good deal below the estimated trend, and below other recent polls. In part this reflects a typical Harris "house effect". With it's atypical four point question format ("Excellent, good, fair or poor" rather than "approve or disapprove"), Harris' approval results are often a bit below that of other pollsters. The plot below, however, shows that this low result is not only due to house effects. The current poll is noticeably further below the trend estimate than are other recent Harris results.
How far the new poll is from the trend is apparent in the next figure below, in which Harris clearly falls outside the 95% confidence interval for "normal" variability. While it is possible the new poll at 28% represents a sudden negative shift, this is quite a distance away from other contemporaneous results from CBS News, NBC/WSJ and Pew.
In light of the outlier analysis, it seems likely that the new Harris poll does not represent a sudden shift in support for President Bush. It is more reasonable that our prior estimate of 34.7% is closer to the national mood, rather than the 34.1% resulting from inclusion of this poll. In either case, the overall picture of relatively little trend in the last 5 months holds. We have been in a period of approval moving up or down a point, but never establishing strong evidence for a trend. While it could change at any point, the post-2006 election period has so far held in rough equilibrium.
Cross-posted at Political Arithmetik.