Approval of President Bush's handling of his job now stands at 38-39% according to the model trend in the above graph. (Hint: Click on graph, then click again for optimal image quality.)
New polls from Fox (Approve 39/Disapprove 54), Gallup (38/60), LATimes/Bloomberg (38/58) and Quinnipiac (36/58) confirm the decline in President Bush's approval rating over the past 2-3 weeks. They also confirm the downturn that CBS News/New York Times polling found earlier, though none of the new polls match the 34% approval in the CBS/NYT poll. The result of combining the polls, however, is to remove any doubt that CBS was an outlier. The newly estimated trend line in the graph above is virtually unchanged whether CBS is included or not in the calculations. With the new data, the estimated approval based on my model is now 38.4% with the latest CBS poll included, and 38.9 without it.
The red line for the trend without CBS is considerably closer to the overall trend than it was earlier in the week before new data arrived. (See the previous graph and discussion here.) Even then the data did not suggest that CBS was an outlier relative to normal variation in polls. With the new data, the CBS poll is even less distant from the predicted value of the model.
In short, the new data confirm the trend that CBS indicated. The tendency of CBS to exhibit a "house effect" of -3.028% points lower approval of President Bush is also consistent with a trend of 38-39% approval. The CBS estimate of 34% plus the house effect of 3% would say that CBS adjusted for house effect predicts a model approval rating of 37%, within 1.4% of the model estimate. Since the model has a margin of error of 3.8%, this is well within the range we would expect.
