The Gallup Poll

The Gallup Poll

When Gallup conducts a national opinion poll, the starting place is where all or most Americans are equally likely to be found. That place is in their home, which is the starting place for nearly all national polling. The actual target audiences, referred to as "national adults", are aged 18 and over, living in telephone households within the United States. What I don't understand is that Gallup excludes college students living on campus, armed forces living on military bases, prisoners, hospitalized people, and anyone else who is living institutionalized. I think these exclusions are unfair. The article explains that the reasoning for not including the people who live in the places mentioned above is because of the difficulty in reaching the institutionalized population. The article claims that this compromise is necessary, but I think they are over-looking people that have every right and to participate in a sampling of American "national adults".

The main objective of the Gallup Poll is to give every American household, and every American an equal chance of falling into the sample. Since nearly 95% of all households have a telephone these days, surveys are conducted by telephone. This makes a lot more sense and is more convenient than having to go door to door to interview people as they did in the past. I couldn't even imagine how annoying that would be.

Now, I find that the design method used to sample the Gallup Poll's target population randomly is very interesting, and seems very effective in ensuring that the sample is absolutely random. They start with a computerized list of all telephone exchanges in America, as well as with estimates of residential households these exchanges have connected to them. A process named random digit dialing (RDD) is used, where a computer creates phone numbers from the exchanges, then generates samples of telephone numbers from those. What this means is that this process creates a list of all possible telephone numbers a household could have, then picks a subset of numbers from that list for Gallup to call.

About 30% of residential telephone numbers are estimated to be unlisted. This is why Gallup uses this method instead of just selecting numbers at random from the telephone book. The article states that if they did in fact select from the telephone book at random, missing the approximate 30% of unlisted numbers, a possible bias would be introduced into the sample. I don't understand why Gallup is concerned about a possible bias occurring in the event of leaving out some unlisted numbers, but when it comes to excluding Americans who are living institutionalized, the possibility of a bias there is left unmentioned. I think it is just as biased, if not more so.

I do think that to a large degree, this practice that Gallup uses for random sampling is sound. The procedure that they use in selecting telephone numbers seems to me, to be completely foolproof, and the best method possible in order to ensure an unbiased and completely random sample. But again, this system only reaches Americans in residential households. The number one question I can't get out of my head is: What is the percentage of "institutionalized" Americans that do not have an equal chance of falling into the sample?

The typical sample size of a Gallup Poll is 1,000 people, which is meant to be representative of the United States actual population of 187 million. At first I thought this was totally unreasonable. How can only 1,000 people be the voice for 187 million? This article explains why Gallup uses this sample size well enough that makes me think that the survey of only 1,000 people is a reasonable number.

The Gallup's justification for using such a seemingly small number has a couple of different factors. The article explains the obvious, that a sample from 1,000 people is going to have more accuracy than a sample from twenty, but also goes on to explain that once the sample gets to a size of 500-700 or more, the accuracy gains that come from increasing the sample size becomes less and less.

Another thing I found interesting in this article is that Gallup as well as other major organizations use sample sizes of 1,000-1,500 because these sample sizes are enough to give a good balance of accuracy opposed to much larger samples which would be very expensive. I hadn't considered expense when it comes to collecting a sample but apparently it can be very costly. For instance, if a sample size of 4,000 were selected each time there was a poll, the increase in accuracy would be very small, and would not justify the increase in cost. Now, I believe this to be true because I do think that the people behind things like the Gallop Poll are very smart statisticians and would not lie when it comes to what effects the sample size has on