Thursday, January 18, 2007

The Dynamics of Reaction to Bush's Surge Speech

(Click on the figure for a full resolution view.)

Shifts in opinion on Iraq immediately following President Bush's speech last week were not as uniformly negative as some reporting implied. However, the response turned more negative as time passed following the speech. The immediate aftermath of the speech found generally small shifts, yet more of these moved towards the administration's positions than against them. In contrast, by the weekend and later polls found most opinion moving against the administration. This difference reflects the role of congressional reaction and media coverage of that reaction in shaping response over time, in contrast to the immediate response to the speech itself.

There are now five polls conducted entirely after President Bush's speech on Iraq last week. (CBS News did a "call back" with previously interviewed respondents the night of the speech. I'm leaving that out because it isn't a fresh sample and not all original respondents could be reached that night.) ABC/Washington Post did a post-speech survey on January 10. CNN/Opinion Research Corporation (CNN/ORC) conducted a poll on 1/11. The Pew Research Center conducted interviews 1/10-15. USAToday/Gallup polled 1/12-14. LATimes/Bloomberg interviewed 1/13-16.

While partially overlapping, these polls are ordered in time and let us look at how reaction to the speech has developed over time. The Pew survey had the longest field period and completely overlaps the USAToday/Gallup poll. Pew has two days of interviews before Gallup began and one day after Gallup. I'm assuming this means the average Pew respondent was interviewed earlier than the average Gallup respondent. Polls often (but not always) complete most of their interviews early in the interviewing period, which would push the average Pew date earlier as well, but I don't know if that was true of this specific Pew poll. If this assumption is false, then my story below is not quite so neat so bear this caveat in mind.

There are a large number of questions that have been asked about Iraq in these polls. Each reflects support or opposition to some aspect of the administration's policies, approval or disapproval of the President's handling of Iraq, and optimism or pessimism about the prospects for success in Iraq. Usually we focus on a single question such as support or opposition for the troop increase. But we can also get a sense of how the wider range of opinion is shifting by looking at a number of questions simultaneously. Here I take the latter approach. The data I use are all questions in any of the five surveys for which there is a "before" and "after" reading of opinion. Most of the "before" readings are from January or December, though a few come from earlier. The specific questions are listed at the bottom of this post.

The figure above shows how opinion shifted from before to after the speech for all these items. The arrows begin at the "before" and point to the "after" poll result. In this figure I don't try to label the specific questions (I do that below.) Here the idea is to look at the overall direction of movement.

The horizontal axis plots negative opinions about the war or the administration, while the vertical axis plots positive views. The closer to the diagonal line a poll is the more people expressed a positive or negative opinion, while the closer to the lower left corner the more people said they didn't have an opinion or adopted a neutral position (such as the increase in troops won't affect the outcome either way.)

Arrows that point toward the upper left corner indicate movement that favors the administration's positions, moves toward a more positive view of the president (or a negative view towards the Democrats) or expresses increased optimism (or reduced pessimism) toward the war or its likely outcome. So arrows pointing "northwest" (or due west) are good for the administration. Arrows pointing the opposite direction, southeast or due south, are the opposite: bad news for the administration. The longer the arrow, the bigger the change. Arrows pointing northeast or southwest are mixed messages with both support and opposition increasing (northeast) or decreasing (southwest). Only a few of these appear in the data.

The arrows are color coded by poll and therefore by time as well. The legend gives the order of polling as well as pollster.

There is a danger here of confounding the effects of time after the speech with "house effects" that reflect persistent differences between polling organizations. I've mitigated that confound here by measuring change within polling organization, from that pollster's "before" to the same pollster's "after" reading. If one organization typically produces higher approval ratings than does another pollster that difference would show up in where the arrows start but the direction and extent of change would be measured from that pollster's baseline, subtracting out the house effect. This isn't foolproof and it is conceivable that house effects also affect change measures, but this approach should mitigate house effects by relying on “apples to apples” comparisons.

Most arrows begin and end in the lower right portion of the figure, reflecting the generally negative balance of opinion on the war. It is this fact that has received the most attention in reports on the post-speech polling. Minorities support the troop increase, approve of Bush's handling of the war, and are optimistic about the outcome of the war. As a result the headlines tended to stress the net negative evaluation while paying less attention to the direction of change following the speech. This leaves the impression that the speech failed to increase support for the administration. If we focus on the arrows, we see a more interesting pattern of change.

Over half of the arrows (20 of 37) point to the northwest or west, indicating increased support for the administration. So the change in opinion was by no means a net negative when viewed across all items and polls. But the shifts are not random. As time goes by, the shifts of opinion become more negative toward the administration's positions. We can see this more clearly by looking at each poll in turn, avoiding the clutter of overlapping arrows above.

The red arrows from ABC/Washington post reflect the most immediate reaction, prior to the influence of media coverage or the impact of congressional reactions. Four of five items in the ABC/WP poll moved in a positive direction. Trusting Bush more than Democrats (+4), approval of Bush's handling of Iraq (+6), and thinking the war has been worth the cost (+4) all moved modestly in the President's direction in this "instant" reading of reaction. To be sure these changes are not huge, but they demonstrate some positive response to the speech. The item that moved more negatively was the belief that we are winning the war (-5). And belief that Iraq is like Vietnam barely changed. To be sure only 36% said they supported the increase in troops in the ABC/WP poll, while 61% opposed. But here we don't have a prior benchmark to know if this was an improvement or not. Certainly given that prior to the speech only 28% in ABC/WP polling approved of Bush's handling of the war, one might conclude that 36% support for the surge was in fact a better than expected showing

The CNN/ORC poll taken the next day found 6 of 9 shifts pointing northwest or west. Those saying Bush has a clear plan on the war increased 10 points in the President's favor. Bush's handling of the war (+1), saying US defeat was not possible (+5), trust that the Iraq government could handle the situation (+1), and trust that the US government can handle the Iraq situation (+8) (though note that this is the US government so this could include increased optimism about a Democratic Congress rather than support for the administration.) And while belief that victory would be the outcome did not increase, those thinking defeat would be the outcome declined (-2). One item failed to change at all, support or opposition to the war overall. And two items moved slightly away from the President's positions: believing that victory is possible (-1) and Bush's overall job approval (-1).

Pew found 6 of 8 items shifting in the President's direction. Saying the US needs more troops (+8), Bush has a clear plan (+3), the US effort is going well (+3), we should keep troops in Iraq until the situation stabilizes (+2), that the Democrats LACK a clear plan for Iraq (+3) and the we should NOT immediately withdraw (+2).

Only on the question if whether the war was a mistake did opinion move away from the administration (-2). Those favoring a timetable for withdrawal moved ambiguously with both positions increasing slightly.

So the initial polling was quite consistent in finding small movements in favor of the administration's positions. Only a few of these shifts were outside the margin of error, but the clear preponderance of movement (16 of 22 items) was towards the President's views.

If you scroll back over the figures, you'll note that all of the polls agree on the low level of support for President Bush's surge proposal. The triangular point in each figure locates this amidst the changes. The support for the surge readings are 36%, 32%, 31%, 38% and 36%. These low levels of support provided the basis for headlines that stressed the negative reaction to the President's speech, but overlooked the direction of change in opinion in these initial polls.

As reaction accumulated, the direction of opinion shift turned substantially to the southeast. The blue arrows from USAT/Gallup's weekend poll tell quite a different story from the initial polls. Only 3 of 9 USAT/Gallup items point northwest. Six of the nine arrows point to the southeast, opinions shifting against the administration. Gallup found movement in the President's direction for having a clear plan (+4), Bush job on Iraq (+2) and the Democrats LACKING a clear plan (+9). But for six other items opinion shifted away from the administration: those thinking we will win in Iraq (-3), can achieve our goals (-3), should not withdraw (-1), Bush's overall job (-3), and war has been worth the cost (-2). Those thinking the war was a mistake increased by 1 point though those saying it was not a mistake held steady.

Finally, in the latest poll by the LA Times/Bloomberg only 1 of 6 shifts of opinion favors the president's position, while four move away and one moves ambiguously. The one positive movement is on trusting Bush or the Democrats more on the war. Trust of the Democrats more decreased by 2 points while trust in Bush was unchanged. In contrast the negative movements occurred on items for evaluation of Bush's overall job (-3), war worth the cost (-6), Bush's Iraq job (-2) and believing Iraq is not in a civil war (-5). Belief that we are winning the war moved ambiguously with one point more believing we are winning and one point more saying we are losing (while most people said neither side is winning.)

These changes in response direction are most likely the result of Congressional reaction, including the several Republican Senators who were quite negative towards the troop increase and the related media coverage of the issue. These reactions provided an important element of news reporting that balanced the President's speech with not only predictably negative Democratic reaction but that also stressed the negative reactions of conservative Republicans such as Chuck Hagel and Sam Brownback.

The low support for the troop increase in the polls also played a roll in the developing opinion. By focusing on the low level of support for the surge, news stories about polls emphasized a negative reaction. But they might just as well have focused on items that moved in Bush's direction. These items mostly remained in net-negative territory, so that is also part of the story, but the direction of change should count for something as well in assessing reaction to the speech and the surge proposal.

I do not mean to suggest here that the net opposition to the increase is not part of the story. Far from it. But if our purpose is to assess the impact of the speech, then incremental change is what is most important. Surely no one would imagine that a speech could completely reverse the current views of the war. But if the administration is to improve its standing with the public, it must do so in small, positive, changes. The evidence is that the speech succeeded in accomplishing that modest goal. Where the policy has so far failed is in also mobilizing support from members of Congress. The secondary reaction to the surge policy has overcome the initial movement towards the administration. For my part, the dynamic of opinion shift over the past week is a textbook example of how opinion change depends on both the event itself (the speech) and the subsequent reaction and interpretation.