4.10: Collaborative Activity

The presidential election of 1936 is not the only example of a very public mishap of presidential polling. Probably an even more famous example occurred a dozen years later with the election of 1948. This election was between the incumbent Harry Truman and New York governor Thomas Dewey. Three polls covered the election: Crossley, Gallup, and Roper. As election day neared, all three polls declared that Dewey would win the election. The predicted percentage of the popular vote for each of the four candidates is given in the table below (Freedman et al. 2007; Mosteller et al. 1949).

	Predictions
Candidate	Crossley	Gallup	Roper
Truman	45	44	38
Dewey	50	50	53
Thurmond	2	2	5
Wallace	3	4	4

When election day arrived, Truman received about 50% of the vote while Dewey received only about 45% of the vote. In fact, this error created one of the most memorable photographs in history.

Figure 4.5 shows Harry Truman, the winner of the election, holding a copy of the Chicago Daily Tribune, whose headline declared that "Dewey Defeats Truman.” On election night the press deadline for the paper required the issue of the Tribune to go to press before all the states had reported most of the results. The paper’s veteran political analyst Arthur Sears Henning and the political polls were almost unanimous that Dewey would win by a landslide, and therefore the Tribune went to press with the now famous banner headline (Wendt 1979).

So once again, the polls were wrong, and in this collaborative activity you and your group will discuss the possible reasons for this error by considering doing a poll of your own.

When we discussed the presidential election of 1936, the problem with the survey methods used by the Literary Digest was the fact that the sample of voters was taken from a population that was different from those who voted in the election.

The three survey organizations had all learned the lesson from the 1936 election and attempted to insure that for this election they would sample from the correct population. The method used by all three survey organizations was based on quota sampling. Each interviewer was given very specific demographic information about the people they should interview. This demographic information included categories such as gender, race, and economic status. For example, an interviewer in a certain city might be told to interview seven men. Of these men, three should be under 40 years old, while the remaining four should be over 40 years old. Moreover, one of the men should be African American while the remaining six should be white (Freedman et al. 2007; Stephan and MaCarthy 1958). If the individuals the interviewer spoke with followed the demographic requirements set by the polling agency, the interviewer could speak with whomever they pleased.

1. Suppose you and the members of your group have been hired by the university to conduct a survey about the students’ opinions of the meal choices on campus. You are to interview twenty students according to the following demographics:
   1. Five students should be first-year students, four should be second-year, four should be third-year, three should be fourth-year, and four should be graduate students.
   2. Four of the first-year students should live in the dorms, while the remaining first-year student should live off campus.
   3. Two of the second-year students should live in the dorms, while the remaining second-year students should live off campus.
   4. One of the third-year students should live in the dorms, while the remaining third-year students should live off campus.
   5. One of the fourth-year students should live in the dorms, while the remaining fourth-year students should live off campus.
2. Working with the people in your group, come up with a plan for how the group could find these students and interview them.
3. Considering the plan that your group came up with, are there some problems that you might encounter when implementing the plan? Would some of the students be difficult to find? How could you go about finding these students?

3. Are there some biases that might be inherent in your group plan to find students to interview? Are there parts of the student population that might be difficult to reach? What type of bias could this introduce into the data that you collect?

4. What possible problems with confounding could result from the method your group came up with to find students to interview? Would these confounding problems be important to the conclusions of the study?

5. Now think about the presidential election polls of 1948. Using what you learned from considering the problem of sampling students on campus according using quota sampling, what possible types of bias could have been introduced using quota sampling for the 1948 presidential election poll? Are there members of the voting population who might be difficult to reach? What important confounding problems are there that might have been important to the conclusions of the polls?

6. Keeping the context of this collaborative activity and your answers to the above questions in mind, what is your general opinion of quota sampling? Do you believe that it is a reliable method for conducting polling? What types of problems would one need to consider when implementing this type of strategy?