|
| |
The History of Political Polling
From: Columbia University
| By:
Robert Y. Shapiro |
EDITOR'S INTRODUCTION |
Public-opinion polls were originally designed to give power to the people. In an exclusive interview with Fathom, Robert Shapiro, a Columbia University political scientist and expert on public opinion, traces the history of American political polling to George Gallup, who in the 1930s envisioned his public-opinion surveys as a means of promoting the idea of democracy in an age in which fascism was rapidly gaining ground. |
Fathom: When did political polling begin in the United States? |
 | |
| A 1941 newsreel about George Gallup's American Institute of Public Opinion. | |
Robert Shapiro: Political polling as we know it today began with George Gallup, in 1935 or 1936. Prior to that, there was a long history of what was called straw polling. Before elections, journalists and others would take polls of crowds, either by a show of hands or little written ballots, out of curiosity, to see which candidate was leading in the current election. The practice goes back to the early nineteenth century. |
Fathom: Was there always a connection between polls and journalism? |
Shapiro: The journalistic aspect of polling was always important. Once Gallup's polls became accepted, he would freely issue news releases that the news media would subscribe to, and they would become part and parcel of news stories. Back then, however, there were relatively few pollsters. There were Gallup, Elmo Roper, Archibald Crossley and, later, others like Louis Harris and Daniel Yankelovich. They've been dubbed the public pollsters, in the sense that they would do polls and then issue news releases to subscribers. Back then, the news media would not do their own polls, and the mass media would widely report the polls that were done by the public pollsters. |
Fathom: What was unique about Gallup? |
Shapiro: Gallup began doing his national surveys in 1935, using what would have been called the best available scientific research methods. They involved the use of statistical sampling, not statistical sampling in its purest form, known in the textbooks as simple random sampling, but, rather, something approximating that. Gallup would send his interviewers to many randomly selected locations within the United States. Later, once he got his operation going, he would typically select and use 300 sites. He would send his interviewers to those sites and then instruct them to fill certain demographic population characteristic quotas in the samples. |
The critical underlying concept in doing national surveys is to get representative samples. It was important that every eligible member of the electorate or member of the public would, in the beginning, at least, have an equal chance of being selected. That is, you wouldn't allow for self-selection. The researcher would randomly select, as in a drawing from a lottery, in which every number has an equal chance of being selected. |
Because of the expenses involved in doing that the right way, which would require enumerating literally every person in the United States and then randomly selecting a person, the best that researchers could do during that time was to sample in multiple stages. First, they'd select a location, and then have interviewers attempt to randomly select their respondents. That was considered far superior to methods of straw polling, which involved simply haphazard, convenient kinds of samples of voters. And it was also superior to the most visible straw poll of the day, which was the poll conducted by the Literary Digest, a popular magazine during the 1920s. |
Fathom: What was the Literary Digest poll? |
Shapiro: In each presidential election year in the 1920s Literary Digest would send out ballots to all its subscribers, and also to lists of people with driver's licenses and to other lists that were available, asking whom they would vote for in the upcoming election. Up until 1936 the Literary Digest straw poll was extremely accurate in predicting not only the winner of the election but also the margin of victory.
In 1936 Gallup challenged Literary Digest, stating that the Literary Digest methods were flawed because it was a straw poll, involving self-selection, and it didn't have the quality of random sampling and "representativeness" inherently built into it. In 1936 we had what was called the battle of the pollsters. |
In 1936, the Literary Digest poll result predicted that Alfred Landon would defeat Franklin Roosevelt, in contrast to the Gallup poll, which predicted that Roosevelt would beat Landon. It turned out, of course, that Roosevelt beat Landon. Literary Digest wound up with egg on its face, and not long after that it went out of business. |
What's not commonly appreciated is that even though Gallup predicted the winner, Gallup also underestimated the size of the Roosevelt landslide. So that was a signal that his method was superior to the Literary Digest straw poll method but that there were perhaps problems with the Gallup poll methods, too. |
Fathom: What motivated Gallup? |
Shapiro: Gallup was trained as a psychologist. He had a Ph.D. in psychology, and he was evidently interested in using the latest social-scientific survey methods to study, in particular, things related to attitudes and psychology. It was almost natural--and not surprising--for him to get involved in public attitudinal research, public-opinion research and market-related research. |
The political polls that Gallup did were not his primary source of livelihood. That is, the polls were done to draw attention to his public-opinion research methods and to the potential applicability of those methods to studying other aspects of public attitudes, in particular, things related to market research. And the company continues doing those things, although it may make more money today from doing different kinds of political polling directly for the mass media, with magazines and other sources. |
This whole battle of the pollsters was primarily an advertising vehicle for the Gallup organization, and it was very successful. Gallup became the father of public-opinion polling as we know it today, and had a lot of influence, with Gallup methods and polls popping up in different countries around the world. His success also promoted the future success of market research and public-opinion research for profit, rather than in the public interest. |
Fathom: What was Gallup's greater purpose in doing polling? |
Shapiro: Gallup talked about public-opinion polling from the standpoint of being a populist or democrat. That's democrat with a little "D," not a Democrat or Republican partisan. He thought public-opinion polling was a way of promoting democracy, a way of channeling the opinions and preferences of the public voters into the electoral process between elections. The elections themselves became indicators of what the public wanted in terms of preferred candidates and parties, but polling became a way of targeting, of fine-tuning information about public opinion, finding out where the public stood on different kinds of issues. So the polls were a vehicle for promoting democracy. |
Gallup's polling took off during the 1930s and 1940s, following developments in Europe including the Russian Revolution and the rise of Communism as well as the rise of fascism in Italy and Germany and Spain. So there were various forms of government out there that posed significant challenges to American democracy as we knew it back then, or liberal democracy as we hoped it would be. |
During that period, there were all kinds of debates about which direction the world would go, in terms of the furtherance of democracy versus the rise of Communism or fascism, following a time in which the dominant forms of government were monarchies and empires. Polling became a way of promoting the idea of democracy and, perhaps, taking it one step further, of promoting the virtues of power to the people and contrasting it with more authoritarian forms of government that were developing in Europe. |
Fathom: What kinds of changes have occurred in the methods of political polling since 1936? |
Shapiro: One of the problems with the kinds of polling that Gallup did, which wasn't fully appreciated during that period, given the success and the ostensible accuracy of the polls, was the degree of discretion granted to interviewers with regard to whom they would interview. They were given demographic quotas to fill--age and sex distributions of members of the public that they were supposed to talk to. That introduced a certain self-selection that could conceivably bias the polls. |
This bias became fully appreciated in 1948, another watershed year. It was a year that could be called "the year pollsters put their feet in their mouths" or "the year the pollsters did themselves in." The famous event was the re-election of Harry Truman, who defeated Thomas Dewey in the face of predictions, weeks before the election, that Dewey would win. There's the famous photograph of Truman with a prematurely printed copy of the Chicago Tribune that reads, "Dewey Beats Truman." This episode brought to light some of the problems in the methods of the pollsters. |
One problem that occurred was that pollsters, out of either arrogance or just misjudgment, thought that once the voters had established their support for the candidates, as early as Labor Day, nothing would change. There was a common wisdom that, after Labor Day, nothing changes. So the pollsters stopped polling two weeks or more before the election. And in the period between the time they stopped polling and the election itself, public opinion evidently changed. |
Fathom: What was the outcome of the 1948 election in terms of the pollsters? |
Shapiro: After the 1948 election, there was an outcry about the problems in polling. The Social Science Research Council commissioned a study that resulted in a book about the 1948 election and the problems of polling. The pollsters then did some serious soul-searching and research about what the problems were. Beyond the obvious fact that they stopped polling too early, what became apparent was that their methods were flawed. |
The pollsters from then on refined their methods in ways that gave interviewers less discretion. Gallup would still send his interviewers out to 300 locations, but they would be given more systematic instructions with regard to selection of households and selection of respondents, in ways that would not give them as much discretion as they had before. |
They couldn't simply go to shopping centers or, as could be done during the New Deal period, interview members of WPA gangs that were taking lunch breaks. They couldn't simply gravitate toward friendly locations. From then on, they were required to actually go to street blocks, from which point they were given instructions to start with a certain household in the beginning, moving in one direction or another from that household until they were able to fill their quotas. They were given instructions to go to households that they might not have normally selected, based on the appearance of the household, the likelihood of finding somebody at home or, more important, of finding somebody who would be cooperative and who would talk to them and complete their interview. |
Fathom: And did this change of method lead to greater accuracy? |
Shapiro: In terms of the reliability and accuracy of the polls, we haven't had a fiasco like we had in 1948. At times there's some debate about how the pollsters could be more accurate. But, clearly, the Gallup polls and the other polls have been closer to the mark in the elections, as a result of fine-tuning their methods. |
Fathom: How have methods been fine-tuned over the years? |
Shapiro: Some surveys, particularly later academic surveys, were done in ways that actually gave the interviewers even less discretion in terms of whom to interview. Rather than filling quotas, the surveys used probabilistic methods for randomly sampling people within households. That is, the interviewer would go to a household and, rather than interviewing the first person he or she encountered, the interviewer would do things like ask the first person contacted in the household to provide a list of everybody in the household. Then the interviewer would randomly select, by means of a table or some other means, which person in the household to interview. This way the person in the house was randomly selected, and not selected based on a quota or any discretion that the interviewer might have. So there was fine-tuning of that sort; these weren't watersheds, but they indicated the kinds of methodological detail that researchers were attentive to. |
Fathom: What technology had the greatest effect on polling? |
Shapiro: In the 1970s, survey researchers determined that it was possible to do effective polling by telephone rather than through in-person interviews. That dramatically lowered the cost of doing surveys, and it provided more incentives for doing national surveys rather than surveys of localities. That is, in doing an in-person survey, researchers oftentimes, in order to cut costs, might focus on a specific town or locality or state. |
With the advent of the use of telephones, surveys could be done much more cheaply, promoting an increasing number of national surveys. Studies determined that telephone surveys produced results comparable to face-to-face surveys, and led pollsters to say, "Let's drop the face-to-face surveys and go with the cheaper method." Also, it turned out that, from a scientific standpoint, telephone surveys perhaps represented a closer approximation to the idea of a simple random survey, in which one began with the equivalent of a list of everyone in the United States, that is, every telephone in the United States. |
The way the samples were drawn was not to use lists of telephone numbers but to use a method known as random digit dialing. There are different variations of the method, but, in each instance, what's done is that random numbers are made up, so that all combinations of numbers become eligible. People who had unlisted phone numbers became possible targets using that method. To the extent that most households had telephones, this allowed for the possibility of doing representative random sampling of all households in the United States. |
Of course, not every household had a telephone. That is, the penetration of telephones, around the time they started, was on the order of 90 percent or more; now it's up at 95, 96, 97 percent. The problem was that in certain geographic areas--in particular, in poor rural areas in the South--the availability of telephones in some localities, when they started doing these telephone surveys, was on the order of 70 percent rather than 90 percent. So there were households whose voices were locked out of the process. On the other hand, the trade-offs in terms of costs and benefits were so overwhelming that, by the end of the 1970s, Gallup and other organizations had shifted over to telephone interviews. |
Also, at that time, as polling became less expensive, the news media themselves were able to get involved in polling, for reasons having to do with the fact that they did not want to be dependent on the public pollsters like Gallup and others. More important, they did not want to be at the mercy of political parties and candidates, who began doing their own polling and could then issue their own press releases on the state of public opinion. |
Fathom: What role, if any, will the Internet have in affecting future political polling? |
Shapiro: The idea of shifting from telephone polling to Internet polling is a lively topic of debate at the moment. The one new juncture that survey researchers have reached is that, with the success of telephone polling, there's been a proliferation of telephone polls. As a result of that, and as a result of telephone solicitations and market research, as opposed to political or academic polling, a lot of people, when they're called by telephone, decline to participate, because they think the caller is trying to sell them something or because they're just being bothered too heavily by all these phone calls. |
In order to do good polling, because people are less willing to participate, more calls and callbacks to households have to be made, and as a result telephone polls are becoming more expensive. Therefore, researchers have been thinking of shifting from doing polls by telephone to doing polls over the Internet. The Internet also cuts costs even further, because it does away with interviewers. With polling by the Internet, when people type in or click in their responses, the data go directly into a computer data file, bypassing other old-style intermediary stages in survey data processing. This is an additional cost-saving aspect of Internet polling. |
Internet polling also allows people to participate in polls at their convenience, rather than when the interviewer calls. They can be invited to a website, which they can visit at their leisure, or be sent a questionnaire in an e-mail message, which they can respond to at their convenience. They don't have to worry about being bothered and disrupted in the middle of dinner. That's the advantage of doing Internet polling.
The disadvantage has to do with the proportion of the population that has access to the Internet. But, as the number of people who have access to the Internet grows, the possibilities of doing Internet polling increase--provided that people can be enticed to participate in these polls. |
There are two methods that are being used today to develop samples through the Internet. One method is the one that's being pursued by a company called Harris Interactive. What they have attempted to do is entice people who are on the Internet, by using displays and promotions. For example, people who might go to a soap-opera website might see an icon that said, "Visit our Harris Interactive website." If they clicked on that, they would be sent to the website and asked to allow their e-mail address to be part of Harris Interactive's data bank, their survey population. Later on, a random sample of people from this data bank of e-mail addresses would be selected, contacted and invited to participate in a survey. |
The disadvantage of that method is that you're not reaching people who don't have access to the Internet, even if you have 30 million people who provide their e-mail addresses. But another company called Intersurvey is confronting this problem by using normal survey methods to draw a sample of people, and then providing Web access to those people who don't have it. Internet polling has big advantages for market research, and a lot of people in survey research think the Internet is part of the wave of the future, assuming that, at some point in time, the penetration of computers will be sufficiently widespread to allow representative samples to be drawn. A lot of people are exploring it. The next election will provide a setting for comparisons between Internet polls and telephone surveys. |
|
| |
