10 -- Part 7: THE AMERICAN LEGAL SYSTEM AND THE COURTS
A sample is the portion of the population that a politician or pollster surveys.
The sample is used to make an estimation of what everyone else thinks.
If the sample is chosen to be representative of the whole population, it works very well.
A truly representative sample is one that does not overrepresent any part of the population and whose responses can be generalized to the whole.
When a sample is not chosen scientifically and has too many people in it from one part of the population, we say it has a problem of to judge public opinion from what they hear among their supporters and friendly interest groups.
They will get a misleading idea of public opinion if they don't know how those they meet differ from the full public.
The name for these polls comes from the fact that a straw is thrown up into the air to indicate how wind is blowing.
One of the biggest mistakes in Americanjournalism was made by Harry Truman.
It's not likely that similar goofs will happen with polls all the way up to Election Day.
Polling is a relatively precise science.
Political polls are just a small part of the marketing business, which spends a lot of money trying to gauge what people want and are willing to buy.
Many local governments conduct surveys to find out what their citizens want and how satisfied they are with municipal services.
All polls face the same two challenges, one of which is getting a good sample, which entails both sampling the right number of people and eliminating sample bias, and another of which is asking questions that yield valid results.
The sample should be close to the population it is drawn from.
It might seem counterintuitive, but a sample of only a few hundred people can be used to represent the entire United States with more than 300 million residents.
A poll's reliability is indicated by its sampling error, which is a number that tells within what range the actual opinion of the whole population would fall.
The "margin of error" is usually plus or minus 3 percent.
There is a 95 percent chance that the real figure for the whole population is within 3 percent.
There is a 95 percent chance that between 57 and 63 percent of the population approves of the president's job performance when a poll reports a presidential approval rating of 60 percent and a 3 percent margin of error.
A poll that shows one candidate leading another by 2 percent of the projected vote is too close to call since the 2 percent might be due to sampling error.
The larger the sample, the smaller the sampling error, but the bigger the reliability problem.
It costs more and takes more time to survey five thousand people.
Everyone should have the same chance to be interviewed.
It is possible to get a representative sample in telephone polls since almost all households now have telephones.
Pollsters say that respondents are more open and cooperative when they are interviewed in person.
It's much harder to get a representative sample for in-person interviewing than it is for in-person interviewing.
Cell phone polling is more difficult for pollsters since they are not allowed to autodial them, as people rely on their cell phones more than landlines.
Sampling bias is not a problem that plagues modern pollsters, but there is a way it can sneak in through the back door.
Sampling bias occurs when the opinions of those who choose to participate in a survey differ from the opinions of those who don't.
Over the years, response rates to telephone surveys have dropped, sometimes as few as one-quarter of those intended to be included in surveys actually participate.
The reasons for this drop include hostility to telemarketers, the increasing use of caller ID, and the fact that people are working more, and have less time and inclination to talk to strangers on the phone.
One consequence of the nonresponse problem is that the most reluctant respondents are more likely to be missed in a typical survey than the average population, meaning that a standard survey might yield responses that are slightly more liberal on racial matters than might be the case.
During the analysis of the results, under- or overrepresented groups are compared to their actual numbers in the population.
Studies of differential response rates, which one might think would cause serious sample biases, found that well-constructed telephone polls still provide accurate information on citizens' responses to most questions about politics and issues.
Technology can also create unexpected challenges.
The substitution of computers for humans to do telephone interviewing happened with the advent of computer technology.
The computers dial the numbers and deliver recorded messages, even "Interacted" by asking questions that are answered by pushing buttons on a touch-tone phone.
robo calling is cheaper than using human interviewers, but it is also controversial.
It is easy to abuse, especially when combined with push poll methods.Legitimate polling firms also userobo calls and have collected more information on more political subjects than has been available in the past, such as the state-by-state results provided by SurveyUSA.
We don't mean the polls that CNN or others put up asking for volunteers to click in their opinions.
Pollsters create panels of internet users who regularly log in to deliver their opinions.
Proponents argue that the Internet polls match results from traditional telephone interviewing because they do not rely on strict probability samples.
The advantage of online surveys is that they have the advantage of getting fewer refusals, and they seem to be more candid in admitting to things that might be embarrassing to confess to a human interviewer.
Forbidding autodialing is one of the technologies that can't be used to contact cell phone users.
People are more likely to refuse to answer polls if they are contacted by cell phones.