Mirror Sites:
CSUDH - Habermas - UWP

California State University, Dominguez Hills
University of Wisconsin, Parkside
Soka University Japan - Transcend Art and Peace
Created: February 16, 2002
Latest Update: February 16, 2002
[email protected]
[email protected]

Getting a Sample Isn't Always Easy

Copyright: Jeanne Curran and Susan R. Takata and Individaul Authors, February 2002.
"Fair use" encouraged.

When Measures Reflect Class and/or Agency

The Chicago Tribune Headline, 1948 The Famous Headline.
Election Surprises: Truman's 1948 Victory By John S. Cooper. Published on: December 10, 1999. Scroll about two-thirds of the way down the file to:

"But some people saw the signs correctly, even if they didn't believe what they saw. In Kansas City, a feed supply company conducted their own informal poll by putting donkeys and elephants on the feed sacks, giving farmers the chance to register the preferences by which sacks they purchased. By early September, 20,000 farmers had been polled this way and 54% of them had voted Democratic. The company abandoned the survey saying, "We read the Gallup and Roper polls that were all for Dewey and we decided that our results were too improbable." A similar "popcorn poll" at movie theaters netted the same improbable results.

"Just before the election, Gallup gave the election to Dewey by 49.5% to 44.5%. Crossley said almost exactly the same. Roper gave Dewey 52.2% to 37.1% for Truman. When the election was over, Truman had beaten Dewey by more than two million votes and carried 28 states (303 electoral votes) to Dewey's 16 (189 electoral votes). On the way back to Washington, Truman stopped in St. Louis and held up a copy of the Chicago Tribune (published too early) announcing in a big headline, "DEWEY DEFEATS TRUMAN." He had a great deal of fun imitating national commentator H.V. Kaltenborn's staccato-pitched voice predicting his defeat in election night broadcasts."

Popcorn polls and feedsack choice polls might give a better measure than telephones. They did, in 1948. In 1948, the group that voted Truman in was less likely to be included in a telephone poll.

Social scientists use probability samples when they have access to a population. So that you would take a probability sample for a national telephone survey because you could get number lists and randomize in accordance with what you're looking for. For example, you could take a given proportion of males, females, head of household, age groups, etc. Then you could randomly generate the numbers. Probability sample.

When Agency and Structural Context Put Your Subjects Out of Reach

But there are problems with that, as election polls have found, that the use of telephone surveys doesn't include lots of people who don't have telephones. That can lead to disastrous results, as it did in the Dewey-Truman election in 1948. So you want to be sure you can really access your population with the techniques and methods you're using.

A tempting present-day example is the temptation to use postal codes. But if you're working with illegal immigrants, the homeless, young non-working populations like street kids, resident addresses are of little help in locating your population. Witness the problems of accurate immigrant and ethnic group counts with the census.

In such a case, you need to build a separate sampling frame. That means you go out and canvas everywhere the group you're trying to study might be, include each place identified in your sampling frame, and then sample randomly from your sampling frame. Much more complex and expensive than using residential addresses to locate your sample.

The expense and time constraints often lead to techniques like "snowball sampling." If you're working on a thesis, and you have no funds available to create a sampling frame of where street kids might be found, snowball sampling, a nonprobability technique, means that you grab the first subject you can find who fits your description, and that subject will lead you to two or three more, each of whom will lead you to two or three more, and so on. We can't count on randomness. But we can get a sample. No evidence that the sample is "representative" of street kids as a whole, but if we persist through collecting data in many of the places where street kids hang out, our faith in the generalizability of our results should increase. See Lecture Notes - Sampling Professor Thomas P. Novak, Marketing Research 461, OGSM, Vanderbilt University.

Nonprobability samples are usually used when data must be collected on the fly, while focus remains on reaching those we are sampling: in other words, a convenience sample. For some clear discussion of this methodology see Abundant Fields, Meager Shelter: Findings From a Survey of Farmworker Housing in the Eastern Migrant Stream � Housing Assistance Council, 2000.

---

---