08/30/2008 11:35 pm ET Updated May 25, 2011

"Bounce" Update

I spent much of yesterday in taxis, airports and planes, and most of the last 24 hours sleeping and reintroducing myself to my family after a five day absence. So my apologies for being away when many of your are curious about the "bounce" and reactions to Republican VP choice Sarah Palin.

Let's start with "the bounce." The Gallup Daily tracking survey conducted Wednesday through Friday shows Barack Obama leading John McCain by eight points (49% to 41%) up from the 45-45% tie measured the full week before the convention. The Rasmussen Reports automated survey conducted over the same three nights shows Obama with a four point lead (49% to 45%), after they had him "leading by just one or two points for most of August."

The most important question, tossed my way by my colleague Marc Ambinder, is whether this shift represents a momentary "spike" or a real and persistent change in voter preferences. And the short but frustrating answer is, there is simply no way to know for certain right now.

One problem is that we have a very limited sample of past convention "bounces" to examine -- a sample size of 10 elections and 20 conventions since 1968 -- and the patterns within that sample have been inconsistent. Some bounces persist, some fade almost immediately.

Another problem, which I explored in a column earlier in the week, is that almost all of the "bounces" of conventions past were measured by multi-day polls conducted a week after the convention without another convention underway. This time, the virtual overlap of the two conventions prevents us from obtaining post DNC "bounce" numbers that are comparable to past measurements. So my recommendation is to avoid historical comparisons.

Finally, consider a few reasons why conventions bounces are sometimes just momentary blips.

First, most pollsters will tell you that attitudes are likely "in flux" at times like these. Watching one side of an argument for an hour or more might leave an uncertain voter "leaning" in one direction for a few days only to shift back in the other direction after hearing from the other side a few days or weeks later.

Second, the kinds of people likely to have been at home the last few days may have been a bit different than those who were away (and thus not part of the survey). Partisan Democrats and Obama supporters may have made it a point to be at home n order to watch the convention. Also, those who just happened to be home (for whatever reason) were probably more likely to have watched the convention and the news generally than those more likely to have been away (for whatever reason). The combination may result in a momentary "response bias" favor ingof the Democrats.

It is not surprising that the Rasmussen survey shows a smaller shift, since they weight by party identification. If the "bounce" is solely the result of a non-response bias toward Democratic identifiers, party weighting in this situation would make sense. But we don't know that it is. Conventions bounces that are real and persistent often shift party identification and candidate preference. In September 2004, for example, five national surveys conducted entirely after the Republican convention showed increases in Republican identification. Pollsters that weighted by the pre-convention party ID averages would have artificially suppressed the size of Bush's bounce.

So, like it or not, we really won't have a sense of what these shifts mean -- and what they portend for the rest of the campaign -- until the Republican convention ends and the dust starts to settle in about 10 days.

As for the polls out today testing reactions to Sarah Palin, I hope to post more tomorrow, but for now suffice it to say that I put very little stock in one-night quickie polls conducted on the Friday night before a three-day, holiday weekend. The one from Gallup is probably the best of the lot, if only because they provide comparable results from prior one-night quickie polls conducted to evaluate vice-presidential nominees (although most of these were fielded on week-nights). But take it all with huge grains of salt. With someone as unknown nationally as Sarah Palin was just 48 hours ago, first impressions can be exceptionally fleeting.