Pennacchio internal polling shows dead heat in GOP Senate primary

Did Joe Penn help Rick Shaftan write this manifesto?
Does PolitickerNJ.com have any journalistic integrity? If so, read the generally accepted guidelines for releasing polling data and you be the judge:

Back to front page »

August 29, 2006, 3:19 pm
The Times Updates Standards For Use of Poll Results
By Byron Calame

Sunday’s public editor column offering a guide to readers on how to distinguish between good polls and bad ones noted that The Times had recently developed an updated set of standards on polling for editors and reporters. “These standards, coming just as the fall campaign heats up, provide a timely reminder of responsible journalism,” said Jack Rosenthal, the guest columnist.
The comments by Mr. Rosenthal, the president of The New York Times Company Foundation and a senior editor of the paper for 26 years, caused several readers to request copies of the updated standards. Assuming others might also be interested, I got permission from Times editors to post the full report of the committee that developed the standards. (One paragraph listing the direct telephone numbers of editors who can be consulted by reporters has been omitted.)

—-
The following standards were developed by a committee of editors and reporters and should be adhered to when using poll results.
Committee Members:
Charles Blow, Susan Chira, Marjorie Connelly, Paula Dwyer, Janet Elder, Deborah Hofmann, Mike Kagay, Patrick LaForge, David Leonhardt, Rich Meislin, Adam Nagourney, Lad Paul, Merrill Perlman, Megan Thee and Robin Toner
June, 2006
OVERVIEW
Reporting on polls is no different from reporting on any other information we give readers. Polls must be thoroughly vetted, be determined to have been done well, and be free from bias in both the questions asked and the conclusions drawn. Keeping poorly done survey research out of the paper is just as important as getting good survey research into the paper. If we get it wrong, we’ve not only misled our readers, but also damaged our credibility. This holds true for polls on every topic used in every section of the paper.
Survey research published in The Times should meet the standards of the News Surveys department for disclosure, sampling, methodology, reliability and unbiased content. While the use of well-done polls by outside organizations is acceptable, New York Times polls are preferred when both are available.
Our aim in any poll story is to give readers a sense of where the general public stands on the issues of the day and to make note of any shifts in opinion. Sometimes those changes are fleeting, sometimes they are the beginning of a trend and sometimes they are another point on a graph that has already begun a clear trend one way or another.
In recent years, polls have become ubiquitous. Whatever or whomever the product, point of view or candidate, there is a poll in support. There are polls conducted over the telephone, on the Internet, through the mail, in shopping malls and on street corners. Sometimes a computer-generated voice asks the questions; sometimes it is a real person.
Not all polls are created equal and polling can be a tricky business. Poll results can be easily influenced by many factors including the wording and ordering of questions and the way the poll’s respondents are chosen.
DISCLOSURE
Full disclosure is the first measure of research done well. It is the accepted practice of the polling profession to release the full questionnaire, the results (sometimes referred to as the top line) and the methodology. In addition, the sponsor of the poll and the provider of the fieldwork should be known as well. We should not publish any poll unless we have access to those things. If the survey organization is releasing its findings in stages, an exception can be made to the rule requiring release of the full questionnaire. Even then, the organization should release its full methodology and the question wording used for the released data. These requirements also hold for proprietary research.
PRIVATELY SPONSORED POLLS
Polls sponsored by private interest groups must be examined carefully. Even if they are conducted using random sampling, these surveys often include leading or biased questions or the analysis of the data may be deliberately misleading. Sometimes respondents are compensated for their participation. Just as The New York Times does not pay for news, neither is compensating poll respondents acceptable.
REPORTING POLLS IN THE TIMES
Once it has been determined that a poll meets The Times criteria for disclosure, sampling and methodology, it is a good idea to see if there are other polls on the same topic. The News Surveys Department can help with this. The truth often lies in the preponderance of polls’ findings rather than in the results of any one survey. Rarely should a story be built around a single poll finding from a single poll.
Polls used in the paper should adhere to the guidelines spelled out in The Stylebook:
opinion polls. Articles about the findings of a public opinion poll should name the person or group who conducted it, name the sponsor and, if necessary, explain the sponsor’s interest in the subject of the poll. The article should also give the number of people surveyed, the dates of the survey and the procedures used (whether interviews were conducted by mail, by telephone or in person). If the poll studied some group other than the general population — registered voters, say, or married adults — the report should say how the respondents were chosen.
The article should give the probable margin of sampling error for a sample of the size used in the poll, and to aid comprehension it should be explained in a sentence like this: The margin of sampling error for a sample of this size is plus or minus five percentage points, so differences of less than that amount are statistically insignificant. Both the poll’s findings and the margin of error should be rounded to the nearest whole percentage point because results rendered to the tenth of a point suggest an impossible degree of precision.
The terms opinion poll, poll, survey, opinion sample and cross section should be limited to scientific soundings of opinion. They should not be applied to roundups of comment or interviews of people in the street. Indeed, extensive articles of that kind should include a cautionary note that the interviews are not a scientific sampling and that only limited conclusions can be drawn from them.
In addition, if the survey is available electronically, it is desirable to give the reader the web address.
GIVING READERS CONTEXT
Public opinion is often malleable and potentially instantly responsive to major events. In general, we try not to use polls done immediately after an event like a political convention or a debate. So-called - instant polls- are wildly unreliable and The Times rarely conducts them. Depending on the event, it is better to wait a few days, until the public has a chance to absorb the event and develop meaningful opinions. Even then, it is important to remind readers of the context of the poll: “The poll was taken three days after the President announced new tax cuts.”
SAMPLING
In order to represent the population in question statistically, a survey should be based on a probability sample. That means everyone in the population has to have an equal chance of being selected to participate in the survey or at least a known chance of being selected. Statements about sampling error and statistical significance can be used only if the survey is based on a probability sample.
Probability samples include “random digit dialing” (RDD) sampling, which is generally used for telephone surveys, and cluster sampling, used in face-to-face interviewing and exit polls.
A poll in which the respondents are drawn from a list compiled for another reason, like members of an organization or magazine subscribers, is often questionable because such lists are rarely complete or up-to-date. Similar problems are found in surveys based on numbers listed in telephone directories. Not only are telephone directories out of date by the time they are printed, but they exclude unlisted numbers. The entire population does not have an equal chance of being contacted. Non-probability samples are commonly used in Internet polls, call-in polls, intercept polls, blast e-mail polls, etc. Regardless of what a press release may say, there is no valid margin of sampling error for surveys based on non-probability samples.
INTERNET AND OPT-IN POLLS
Self-selected or “opt-in” samples — including Internet, e-mail, fax, call-in, street intercept, and non-probability mail-in samples — do not meet The Times’s standards regardless of the number of people who participate.
Most Internet surveys are based on panels of self-selected respondents. This makes Internet polls problematic. Often the polls have no way of controlling the number of times a person participates. In addition, Internet access is not yet sufficiently widespread or evenly distributed across socio-economic and demographic groups: older people and minority members are less likely to have Internet access. Many people are working on making Internet polls more reflective of the general population and we are open to future developments.
Opt-In surveys are conducted with respondents who volunteer to answer the questions. Some polling companies pay respondents or offer other incentives for people to complete online questionnaires.
In order to be worthy of publication in The Times, a survey must be representative, that is, based on a random sample of respondents.
Any survey that relies on the ability and/or availability of respondents to access the Web and choose whether to participate is not representative and therefore not reliable. The hallmark of any good poll is that the poll taker chooses and pursues the respondent.
INTERACTIVE VOICE RESPONSE
Interactive voice response (IVR) polls (also known as “robo-polls”) employ an automated, recorded voice to call respondents who are asked to answer questions by punching telephone keys. Anyone who can answer the phone and hit the buttons can be counted in the survey – regardless of age. Results of this type of poll are not reliable.
UNDERSTANDING THE MARGIN OF SAMPLING ERROR
The margin of sampling error is the only quantifiable error in a typical random sample telephone poll, but there are other errors too. The refusal rate, question order, interviewer techniques and question wording are all additional sources of error and bias in polls.
A typical nationwide telephone poll of 1,000 respondents has a margin of sampling error of plus or minus three percentage points. This means that in 19 cases out of 20, overall results based on such samples will differ by no more than three percentage points in either direction from what would have been obtained by seeking out all American adults.
The truth is not a random variable, distributed around our poll measure like a bell curve. Rather, a hundred poll measures would be distributed around the truth like a bell curve. If the truth were 40 percent, then our most likely poll result would be 40 percent. Thirty-nine percent and 41 percent would be a little less likely in our poll, and 38 percent and 42 percent less likely still, etc. And 95 out of those 100 separate poll measures would be between 37 percent and 43 percent.
When reporting the comparison between polls, the margin of sampling error of the difference “between” polls must be determined. It is NOT arrived at by adding or averaging the margin of sampling error for the two separate polls. The sponsor of the polls should be able to provide the margin of sampling error between the two polls. If the two polls are each from a different sponsor, check with The Times’s News Surveys Department for the calculation.
REPORTING OF ELECTION POLLS
Polls are as much a part of politics as are votes. Campaigns aggressively push any polls that appear to suggest their candidate is more in step with the majority of Americans than other candidates are. Campaigns are zealous about their candidates’ being cited as the “front runner,” early and often. Media organizations and special interest groups poll constantly during election years. Bad polls often drive the dialogue.
We should avoid using polls conducted by parties or individual campaigns.

HORSE RACE NUMBERS
Extreme caution should be used when citing any poll’s horse race numbers. It is always best to cite more than one poll’s findings. If five polls suggest Mr. Jones is ahead of Mr. Smith, that is more solid than if only one poll has Mr. Jones ahead of Mr. Smith.
Better still to avoid using horse race numbers and instead give readers an impression of a candidate’s strength based on other measures. But if horse race numbers are necessary, the margin of sampling error should be given. The way the race is characterized should be determined in part by how far outside the margin of sampling error the horse race numbers fall.
Keep in mind that the margin of sampling error applies to each number, i.e. both or all candidates. For instance, if Mr. Jones is backed by 50 percent of the voters and Mr. Smith by 45 percent, and the margin of sampling error is plus or minus three percentage points, then simply state that “while Mr. Jones leads Mr. Smith, the difference is within the poll’s margin of sampling error of plus or minus three percentage points for each candidate.” In that case, Mr. Jones could have as little as 47 percent of voters’ support and Mr. Smith could have as much as 48 percent.

LIKELY VOTERS, REGISTERED VOTERS
Poll figures released during campaign cycles are often based on myriad definitions of voters, i.e. likely voters, registered voters, probable voters and leaners. In using any numbers, it is essential to make clear which group of voters is being measured. In addition, when using multiple polls, the reader must be told of any differences. If one set of figures is for registered voters and another for likely voters, the reader must be so informed.
Adding to the confusion of election polling is the amorphous definition of a likely or probable voter. Every polling organization has its own way of defining the voters most likely to turn out on Election Day. They take many factors into account: past voting, current registration, length of residency at current address, etc. The definition of a registered voter is more straightforward, the voter is either registered or not. It is thus best to stick to figures for registered voters when comparing election polls from different organizations.
EXIT POLLS
No exit poll should be used without first checking with News Surveys.
Exit polls are good tools for understanding who voted and what motivated them. But these surveys should not be used to predict the outcomes, no matter how tempting that may be. An exit poll is taken with voters as they leave polling places and lasts all day. Results often fluctuate by time of day.
In the United States, for now, there is only one source for exit polls. The polls are conducted by a consortium of the major television networks along with The Associated Press.
When using an exit poll to report on elections in foreign countries it is essential to make sure that the provider has a history of conducting exit polls in that country. Once again, readers should be given a full understanding of who conducted the poll and how. The story should also make clear how many polling locations there were throughout the country.
COVERING BAD POLLS DURING AN ELECTION
All political campaigns, especially national campaigns, bring a new spate of badly done polls. Many can and should be ignored, but a bad poll can sometimes drive the story and take on a life of its own. When that happens, it may be necessary to report on the poll and its effects. But it is important that the flaws be described in order to provide context. The more we do that, the more transparent our own coverage becomes.

And look what pops up:

Rick Shafton has a great track record of predicting terrific results for his clients and releasing the partial poll:

Here is a link to the poll he “released” for the 2nd Congressional Seat in Nevada;
His client was Sharron Angle.

His release said the following:
“A tracking poll of 306 likely or definite Republicans who plan or who have already voted in the hot Nevada 2nd Congressional District Primary show that Minuteman PAC endorsed candidate Sharron Angle has opened up a 13 point lead over second place candidate Dawn Gibbons.
Angle led the five way contest with 33 percent of the vote, followed by Gibbons at 21 percent and Secretary of State Dean Heller at 19 percent. 27 percent of respondents were undecided.”

Unfortunately for Rick Shafton (or more directly his client) – we have the results of his prediction: Dean Heller won the Primary and Sharron Angle lost with Dawn Gibbons coming in third!

I can only hope Rick Shafton and his Neighborhood Research maintain their stellar track record – if so, we know who wins the June 3rd GOP Primary: Dr. Murray Sabrin!

http://www.thelandofthefree.net/conservativeopinion/2006/08/13/angle-ope...

Rick Shaftan may not have any integrity but at least he is consistent and predictable (pun intended).

for flakes like "Maverick Murray" and "Jersey Joe aka Jersey Joke"?

Republican voters will see through these two juvenile candidates. I'm sure that next Tuesday Zimmer will win by at least 20 points.