Pardons and Public Opinion
I do believe that the buck stops here, that I cannot rely upon public opinion polls to tell me what is right. - Gerald R. Ford, September 8, 1974
Within a week of making the above statement, and pardoning Richard Nixon, Gerald Ford's public approval rating fell from 71 percent to 49 percent. A 1988 New York Times/CBS poll found 64 percent of Americans opposed the idea of a pre-trial pardon for Oliver North. A December 2002 Gallup poll found 54 percent of Americans "disapproved " of George H. W. Bush's pardon of Casper Weinberger. A 2001 Fox News opinion dynamics poll showed 75 percent of Americans thought "justice was denied" when Bill Clinton pardoned Marc Rich. And now, the very latest: a recent Gallup poll showed 67 percent of Americans believe President Bush should not pardon Scooter Libby.
Like most political scientists, I believe scientific polling is a legitimate (and sometimes very impressive) tool for understanding the dynamics and complexities of public opinion. On the other hand, most political scientists agree that many things can go wrong from the point where someone decides to conduct a poll and the interpretation of the final results. Gallup, Roper, Harris and Zogby International occasionally poll "national samples" regarding events and issues which are of little or no concern to many or most Americans. Pollsters actually learned this lesson early on, when they attempted to measure public opinion regarding Franklin Roosevelt's so-called "court packing" plan. The plan was the hot topic in Washington and on the front pages of the nation's leading newspapers for weeks. But pollsters found half of Americans in their samples were completely ignorant of the topic or had no opinion about it. Imagine the frustration of pollsters anxious to sell material sufficient to produce a headline reading, "Majority of Americans Think ..."
It follows that, unless polling results are meant merely for the purpose of entertainment or partisan marketing, scientific measures of the true opinion of the public will use appropriately formatted questions and filter questions so as to guard against 1) measuring false opinion or 2) creating opinion that simply does not exist in any meaningful sense of the language. A pollster can, for example, ask a sample of Americans if they favor, or oppose, N.A.F.T.A. and report responses. There is no law against that style of polling. On the other hand, if one wants to measure the opinion of the public scientifically, it makes much more sense to approach the topic with filter questions that are not simply multiple choice questions: "Have you ever heard of N.A.F.T.A.? What can you tell me about it?" Open-ended questions are clearly more appropriate than - as students call them - "multiple guess" questions. No, you might not get a sexy headline with the second approach. But you are more likely to accurately measure the true opinion of the public.
Imagine a pollster asking these questions:
1. What news stories in the last few weeks are prominent in your mind?
2. (If the Libby trial is even mentioned) Have you generally followed the progress of the trial?
3. Do you know what Mr. Libby was charged with and/or found guilty of?
4. What evidence against Mr. Libby impressed you the most?
5. Would you support / oppose a presidential pardon for Mr. Libby. If so, why? or why not?
The recent publicized Gallup poll did bother to pursue some of these angles, although half-heartedly. For example, instead of allowing respondents to assert the importance of the story on their own - as I have with question 1 - Gallup asserted the importance of the story for them by first asking, "How closely have you been following the news about the recently completed trial of Lewis "Scooter" Libby, the former vice presidential aide?" We will never know how many respondents would not have even mentioned Libby had Gallup pollsters not brought him up.
Instead of allowing respondents to properly demonstrate knowledge of the trial through open-ended questions, Gallup tutored respondents on who Libby was in the first question ("the former vice presidential aid") and why he was in the news in the second: "As you may know, a jury found Libby guilty on four out of five criminal counts ...
Why all of these obvious missteps? Easy. Despite all of the partisan and media interest in the Libby case, a whopping 57 percent of the respondents in the Gallup poll would not even claim to have followed the trial "closely" or "somewhat closely." Thus, 57 percent of respondents were asked, in effect, "Should the president pardon a person we have just brought to your attention that was recently convicted in a trial that you did not follow?"
With respect to the 43 percent who claimed to have followed the trial "closely" or "somewhat closely," no properly formatted questions appear to have followed which would have demonstrated actual knowledge of the trial. Such an approach would not only be more "scientific," it would also allow for more rich, accurate and pertinent - analysis. The Gallup data show, for example, that claims of higher levels of knowledge regarding the trial are clearly associated with higher levels of support for a pardon: "not closely" - 15 percent for a pardon, "somewhat closely" - 25 percent for a pardon, "very closely" - 44 percent for a pardon. Incidentally, the last figure represents a higher level of support than Gallup found among the general public for the pardon of Richard Nixon (38 percent), Oliver North (27 percent) or Casper Weinberger (25 percent).
Within a week of making the above statement, and pardoning Richard Nixon, Gerald Ford's public approval rating fell from 71 percent to 49 percent. A 1988 New York Times/CBS poll found 64 percent of Americans opposed the idea of a pre-trial pardon for Oliver North. A December 2002 Gallup poll found 54 percent of Americans "disapproved " of George H. W. Bush's pardon of Casper Weinberger. A 2001 Fox News opinion dynamics poll showed 75 percent of Americans thought "justice was denied" when Bill Clinton pardoned Marc Rich. And now, the very latest: a recent Gallup poll showed 67 percent of Americans believe President Bush should not pardon Scooter Libby.
Like most political scientists, I believe scientific polling is a legitimate (and sometimes very impressive) tool for understanding the dynamics and complexities of public opinion. On the other hand, most political scientists agree that many things can go wrong from the point where someone decides to conduct a poll and the interpretation of the final results. Gallup, Roper, Harris and Zogby International occasionally poll "national samples" regarding events and issues which are of little or no concern to many or most Americans. Pollsters actually learned this lesson early on, when they attempted to measure public opinion regarding Franklin Roosevelt's so-called "court packing" plan. The plan was the hot topic in Washington and on the front pages of the nation's leading newspapers for weeks. But pollsters found half of Americans in their samples were completely ignorant of the topic or had no opinion about it. Imagine the frustration of pollsters anxious to sell material sufficient to produce a headline reading, "Majority of Americans Think ..."
It follows that, unless polling results are meant merely for the purpose of entertainment or partisan marketing, scientific measures of the true opinion of the public will use appropriately formatted questions and filter questions so as to guard against 1) measuring false opinion or 2) creating opinion that simply does not exist in any meaningful sense of the language. A pollster can, for example, ask a sample of Americans if they favor, or oppose, N.A.F.T.A. and report responses. There is no law against that style of polling. On the other hand, if one wants to measure the opinion of the public scientifically, it makes much more sense to approach the topic with filter questions that are not simply multiple choice questions: "Have you ever heard of N.A.F.T.A.? What can you tell me about it?" Open-ended questions are clearly more appropriate than - as students call them - "multiple guess" questions. No, you might not get a sexy headline with the second approach. But you are more likely to accurately measure the true opinion of the public.
Imagine a pollster asking these questions:
1. What news stories in the last few weeks are prominent in your mind?
2. (If the Libby trial is even mentioned) Have you generally followed the progress of the trial?
3. Do you know what Mr. Libby was charged with and/or found guilty of?
4. What evidence against Mr. Libby impressed you the most?
5. Would you support / oppose a presidential pardon for Mr. Libby. If so, why? or why not?
The recent publicized Gallup poll did bother to pursue some of these angles, although half-heartedly. For example, instead of allowing respondents to assert the importance of the story on their own - as I have with question 1 - Gallup asserted the importance of the story for them by first asking, "How closely have you been following the news about the recently completed trial of Lewis "Scooter" Libby, the former vice presidential aide?" We will never know how many respondents would not have even mentioned Libby had Gallup pollsters not brought him up.
Instead of allowing respondents to properly demonstrate knowledge of the trial through open-ended questions, Gallup tutored respondents on who Libby was in the first question ("the former vice presidential aid") and why he was in the news in the second: "As you may know, a jury found Libby guilty on four out of five criminal counts ...
Why all of these obvious missteps? Easy. Despite all of the partisan and media interest in the Libby case, a whopping 57 percent of the respondents in the Gallup poll would not even claim to have followed the trial "closely" or "somewhat closely." Thus, 57 percent of respondents were asked, in effect, "Should the president pardon a person we have just brought to your attention that was recently convicted in a trial that you did not follow?"
With respect to the 43 percent who claimed to have followed the trial "closely" or "somewhat closely," no properly formatted questions appear to have followed which would have demonstrated actual knowledge of the trial. Such an approach would not only be more "scientific," it would also allow for more rich, accurate and pertinent - analysis. The Gallup data show, for example, that claims of higher levels of knowledge regarding the trial are clearly associated with higher levels of support for a pardon: "not closely" - 15 percent for a pardon, "somewhat closely" - 25 percent for a pardon, "very closely" - 44 percent for a pardon. Incidentally, the last figure represents a higher level of support than Gallup found among the general public for the pardon of Richard Nixon (38 percent), Oliver North (27 percent) or Casper Weinberger (25 percent).
posted by P.S. Ruckman, Jr. | 1:05 AM
<< Home