Statistics and Political Polls

At any given time throughout a political campaign the media may want to know what the public at large thinks about policies or candidates. One solution would be to ask everyone who they would vote for. This would be costly, time consuming and infeasible. Another way to determine voter preference is to use a statistical sample. Rather than ask every voter to state his or her preference in candidates, polling research companies poll a relatively small number of people who their favorite candidate is. The members of the statistical sample help to determine the preferences of the entire population. There are good polls and not so good polls, so it is important to ask the following questions when reading any results.

Who Was Polled?

A candidate makes his or her appeal to the voters because the voters are the ones who cast ballots. Consider the following groups of people:

• Adults
• Registered voters
• Likely voters

To discern the mood of the public any of these groups may be sampled. However, if the intent of the poll is to predict the winner of an election, the sample should be comprised of registered voters or likely voters.

The political composition of the sample sometimes plays a roll in interpreting poll results. A sample consisting entirely of registered Republicans would not be good if someone wanted to ask a question about the electorate at large. Since the electorate rarely breaks into 50% registered Republicans and 50% registered Democrats, even this type of sample may not be the best to use.

When Was the Poll Conducted?

Politics can be fast paced. Within a matter of days an issue arises, alters the political landscape, then is forgotten by most when some new issue surfaces. What people were talking about on Monday sometimes seems to be a distant memory when Friday comes. News runs faster than ever, however good polling takes time to conduct. Major events can take several days to show up in poll results. The dates when a poll was conducted should be noted to determine if current events have had time to affect the numbers of the poll.

What Methods Were Used?

Suppose that Congress is considering a bill that deals with gun control. Read the following two scenarios and ask which is more likely to accurately determine the public sentiment.

• A blog asks its readers to click on a box to show their support of the bill. A total of 5000 participate and there is overwhelming rejection of the bill.
• A polling firm randomly calls 1000 registered voters and asks them about their support of the bill. The firm finds that their respondents are more or less evenly split for and against the bill.

Although the first poll has more respondents, they are self-selected. It is likely that the people who would participate are those who have strong opinions. It could even be that the readers of the blog are very like-minded in their opinions (perhaps it is a blog about hunting). The second sample is random, and an independent party has selected the sample. Even though the first poll has a larger sample size, the second sample would be better.

How Large Is the Sample?

As the discussion above shows, a poll with a larger sample size is not necessarily the better poll. On the other hand, a sample size may be too small to state anything meaningful about public opinion. A random sample of 20 likely voters is too small to determine the direction that the entire U.S. population is leaning on an issue. But how large should the sample be?

Associated with the size of the sample is the margin of error. The larger the sample size, thesmaller the margin of error. Surprisingly, sample sizes as small as 1000 to 2000 are typically used for polls such as Presidential approval, whose margin of error is within a couple of percentage points. The margin of error could be made as small as desired by using a larger sample, however this would require a higher cost to conduct the poll.

Bringing It All Together

The answers to the above questions should help in assessing the accuracy of results in political polls. Not all polls are created equally. Often details are buried in footnotes or omitted entirely in news articles that quote the poll. Be informed on how a poll was designed.