Discrepencies in the National Polls

The Problems with Polls – Other Than the Final One:
As you know, I have been particularly critical of the daily tracking telephone polls, arguing that they fail to conduct enough follow-up calls to reach hard-to-reach respondents. Because Obama supporters are generally harder to reach, particularly the young and the African Americans, overall completion rates, i.e., the percentage of the people called who actually complete a survey, are generally systematically higher for McCain supporters than for Obama supporters. This phenomenon, if uncorrected by accurate weighting, will systematically underestimate the Obama support relative to the support for John McCain.

There is a second problem that is difficult as the completion rates. Pollsters routinely have to estimate the percentage of support for the candidates based on samples that are filtered to those “certain” or “very likely” to vote in the election. Different polling companies use different standards for this filter and they only rarely share the methods that they use. In general, samples that are based more on registered voters, than on the most likely voters, typically report slightly greater proportions of Democrats, but this will depend on the relative motivations of different voting groups to participate.

Current Differences in Polling Results:
These problems with polls are brought into sharp perspective by a very simple exercise with recently conducted polls. There are six current tracking polls reported every day in the Web Site, realclearpolitics.com, which is an excellent source for information that I recommend to any of you who are following this campaign. The tracking polls include polls reported by Rasmussen, Reuters/Zogby, Hot Line, Gallup (regular), Gallup Expanded, and IBD.

The average margin by which Barak Obama leads in these six daily tracking polls is 4.7 percentage points, which makes the contest a relatively close one. This is as of yesterday, October 16th.

At the same time, there have been eight national polls (not tacking polls) completed and released in the past six days. These polls include ones by LA Times/Bloomberg, CBS/New York Times, USA TODAY, Pew Charitable Trust, IPSOS/McClatchy, ABC/Washington Post, FOX, and Newsweek. These are one time polls, not tracking polls, and the callbacks tend to be completed much more thoroughly than with the tracking polls.

The average margin by which Barak Obama leads in these eight one time polls is 9.3 percentage points, which makes the contest a near landslide. These polls are reported between October 9th and the 16th.

The differences between these types of polls are too large to explain away with the slight differences in timing of the polls. As I have said in the past, the tracking polls tend to understate the voters for Barack Obama because his supporters (younger and African American) are simply harder to reach by telephone polling, and the tracking polls do not follow up with sufficient callbacks to harder to reach respondents as successfully or as diligently as the one time polls do.

The Consequence of this Difference:
The national media relies on the national polls, their own and those of their competitors. They routinely, every day, “interpret” the direction of the election using these polls. In this instance, the polls provide two widely different pictures of the status of the election. In the first case, with the tracking polls, the results suggest that Obama is ahead, but that the election remains fairly close.

In the second instance, we have an election that is tending toward a landslide for Barack Obama; a landslide that would pull behind it increased numbers of candidates for most state and national offices, members of House, the Senate and Governors. This is a dramatically different election than the one suggested by the tracking polls, with significant national implications.

The second instance also suggests that Obama’s “race” is diminishing as a factor as the magnitude of the financial crisis seeps into the consciousness of the American voters.

Support for the Probability of a Landslide:
The reason that I looked at this at all is the profound difference in the story that is presented in a detailed analysis of the changes that have taken place among the state polls. Virtually all of the state polls are one time polls, and not tracking polls. Real Clear Politics does a nice job of summarizing these polls by providing an average of the polls conducted in the various states. There are fewer state wide polls in the individual states, and the average tends to overweight the older polls, which are less representative as they age.

The various state wide polls suggest that Barack Obama has enjoyed a dramatic surge in support over the past month that has shifted most of the so-called “toss up” states into the Obama column, shifted states leaning to Obama to solidly for Obama, and shifted some traditional Republican states into the competitive column. This is happening nearly across the Board, suggesting that a near landslide is building in American politics.

There have been changes in most states and these are some examples. Obama is leading in Virginia by an average margin of 8.1 points, and this is a state that a Democrat has not carried in 40 years. Obama now enjoys leads in Ohio (3.8 points), North Carolina (1.2 points), Florida (4.8 points), Missouri (1.8 points), and Nevada (3.0 points). These reflect changes of 8 to 10 points in states with McCain was leading a month ago.

Senator John McCain still retains a very small lead in Indiana and West Virginia, but the result is clearly too close to call. Finally, Georgia and North Dakota, both of which were solidly for McCain, are now in the “leaning McCain” category and attracting Obama resources for the first time.

Implications for Other Races:
I cannot summarize all of the other races that might be affected by the surge for Obama. However, it suggests that the Democrats will clearly win firm control over the Senate, with the outside possibility of reaching 60 seats, and improve their control over the House of Representatives by potentially more than 25 seats.

As I have pointed out in earlier posts, the Obama vote is composed of two segments that have a profound impact on these other races. First, there will most certainly be a dramatic increase in the participation of African Americans, relative to other groups of voters. This is the single most loyal voting bloc in American politics, supporting Democratic candidates at over 90 percent rates in elections from President to the local dog catcher.

The second voting bloc is the new voters under 35, which appear to split more than two to one for Barack Obama. Much more than older voters, these younger voters are likely to vote a “straight party vote” for the Democrats because they are less aware of the other office holders than are older voters.

Both of these voting blocs are much harder to reach by traditional telephone polling. Even with repeated callback attempts, polls frequently undercount these two groups and weighting is required to represent them properly in the results that are reported. Any systematic undercounting in the polls of these two groups will have the affect of diminishing the size of Obama’s margin of victory; thereby making the entire election appear closer than it is in reality. This actually aids Obama because it will stimulate the turnout of his loyalists.

Accuracy in Polling:
Polling companies do not report routinely their true completion rates on their polls. When we were in the process of developing Internet based polling at Harris Interactive, it was clear to everyone in the firm that completion rates in telephone polling were becoming the “dirty little secret” of the telephone polling community.

Completion rates on telephone polls have been falling since the 1960’s, but the problem became acute during the 1990’s with the rise of telemarketing and later with the development of national “Do Not Call” lists. Although telephone polling was specifically exempted from the Federal legislation restricting calls to do not call lists, the general hostility to unsolicited phone calls reduced cooperation and completion rates dramatically. Again, these “refusals”, do not answer; and answering machines, voice mail and the like give respondents easy and inoffensive ways to opt out of telephone polls.

I say “dirty little secret” because the conspiracy of public silence on the issue is a product of the self-interest by the polling companies and their media clients to avoid any real public discussion of the problem. Polling has become a highly integrated part of the news reporting process, and why would either side want to discuss a problem that undercut the validity and accuracy of the so-called “information” they were feeding the public on a regular basis.

The problem is even more acute with the state polls, as most people in the polling community recognize. The budgets for these polls tend to be more limited; the news media sponsors resist paying the cost that is required for accurate polling, involving repeated callbacks; and they have never been as accurate, on average, as the national polls. You can see this yourself simply by reviewing information at the Web Site of the National Council on Public Polls (NCPP).
The problem is compounded because of the candidacy of Barack Obama, where systematic undercounting of African Americans and the younger voters is likely to affect nearly every state, and every race, in which polling is conducted.

The Implications for the Election:
Senator Barack Obama is headed toward a large and perhaps landslide victory, bringing with him most of the close races for the House and the Senate, and even the state legislative races.
He will “seal the deal” with half hour broadcasts later this month that will allow him to represent his best case in his best medium. His “infomercial”, which I have urged for a time, will allow him to present his arguments in the most positive way possible, to what will be a very large audience of people who care passionately about politics, and the infomercial will be discussed by the national media for several days afterward.

We are in a “realigning period” that will reshape American politics for at least a generation. Younger voters, those under 35, are selecting the Democratic Party by a huge margin. These voters, compared with their parents and grandparents, are less religious and more secular, much more activist on global warming and the environment, and much less sympathetic to the political agenda of the Christian Coalition. The “tide” of the influence of the Christian Coalition in American politics is beginning a long process of "ebbing" and we may well be heading toward a politics that sounds more secular in tone.

The “young” have found the perfect candidate in Barack Obama – optimistic by character, moderate in his style and tone, activist in his political orientations, libertarian in his social views, secular in his public policy stances, independent of special interest constituencies, and very, very bright and articulate.

I do not envy Barack Obama! He is inheriting a mess, not of his own making, but that mess will dominate his Presidency for years. Like all of his, he is going to have to learn how to make do with less, and the less will get much worse unless he deals with the entitlement crisis on Medicare and Social Security that is approaching us like a tsunami.

Americans are a very resilient people. With proper leadership, I believe that we (at least a majority of us) will behave properly. Proper leadership, however, means that office-holders in Washington, D.C., and the state capitols are going to have to start telling the truth for a change, even if it makes some voters angry.

The very worse "sin" in a democracy is the systematic lying about conditions and events that all of us can see as lies” eventually. The lying, even when one politely calls it “spin”, erodes all trust, and we can see the terrible cost of the loss of public trust in the current economic crisis. The impact of that loss of trust in the current Administration ought to be apparent to nearly everyone, but will we learn the lesson? That is the question?

Just my opinion,

Gordon Black