In my third year in undergrad studies, I picked up Walter Lippmann’s Public Opinion (1922) and read this opening paragraph:
“There is an island in the ocean where in 1914 a few Englishmen, Frenchmen, and Germans lived. No cable reaches that island, and the British mail steamer comes but once in sixty days. In September it had not yet come, and the islanders were still talking about the latest newspaper which told about the approaching trial of Madame Caillaux for the shooting of Gaston Calmette. It was, therefore, with more than usual eagerness that the whole colony assembled at the quay on a day in mid-September to hear from the captain what the verdict had been. They learned that for over six weeks now those of them who were English and those who were French had been fighting in behalf of the sanctity of treaties against those of them who were Germans.”
I was hooked on studying public opinion…
It presumably began back in 1824 and it was an ominous beginning. That year, the Niles Weekly Register, a newspaper from Baltimore, published the results of a straw poll about a local election (Micheau 2018: 22). The name “straw poll” was inspired by the farmers’ practice to throw straw into the air to determine where the wind blew. Maybe that first published straw poll was inspired by the series of ad hoc attempts throughout the country to try to predict the vote. Around that time, people began to notice counts at meetings and tallies from “poll book” left at taverns and other public places attempting to predict presidential leanings (Bardes and Oldendick 2017: 17–18). Then, during the 1824 US Presidential election campaign, the Harrisburg Pennsylvanian published a poll which wrongly predicted that Andrew Jackson would beat John Quincy Adams (Hoy 1989: 12). Not a great start.
During the presidential elections of 1908 and 1912, the New York Herald teamed up with the Cincinnati Enquirer, the Chicago Record-Herald, the St. Louis Republic, the Boston Globe and the Los Angeles Times to seemingly make erroneous predictions about who would win the Presidency (ibid.: 12). In the 1923 Chicago mayoralty election, the Chicago Tribune tabulated more than 85,000 ballots and released vote preferences by ethnic groups, “streetcar drivers,” moviegoers, and other such small groups despite the lack of application of the most basic sample theory (Frankovic 2012: 114). By the 1928 Presidential election, more than 85 straw polls were published (Micheau 2018: 23) with mixed success. Then came the Literary Digest Poll in the 1936 US Presidential election. The Digest failed to predict Roosevelt’s victory over Alf Landon and the 19.6 percentage point error in the prediction remains the largest such error ever recorded (Lachapelle 1991: 2–3). The Literary Digest folded less than a year later and many would have predicted that the practice of measuring public opinion would soon follow suit. But on the contrary, public opinion polling survived its rough start to emerge as an influential part of the political, business, and social landscape.
In this chapter, I aim to explore four key background areas pertinent to the rest of this book. First, the chapter presents a brief history of polling. This history is by no means exhaustive but identifies a few important signposts. Of particular importance is the discussion of the evolution of polling from its inception to its latest transformation into what we now refer to as market intelligence. This exercise helps contextualize the analysis. I return to more specific historical developments in other chapters to focus on moments which defined the specific strands of opinion research. In particular, I explore the rarely discussed history of commercial opinion research in the third chapter and some recent development in the fourth chapter. But this initial overview provides the overall, relevant context and evolution. Furthermore, this history aims to be less US-centric than the literature tends to be. While there is no denying the importance of American pollsters in establishing and developing this industry, other countries—notably France, the United Kingdom, and Canada—have also made important contributions that are too often overlooked.
Second, a short but much needed discussion about terminology will follow and aim to clarify the way we talk about this topic. Polls, surveys, public opinion research, market intelligence, and polling are too often used interchangeably and this leads to confusion. I define and explain the terms in order to bring some clarity to this discussion. Third, the link between polls and public opinion is reviewed. It is important to be reminded from time to time that a definition of public opinion goes beyond the quantification we are taking for granted. In short, public opinion is more than a sum of individual opinions and this point will be discussed in this chapter. Fourth, some important methodological challenges are reviewed. The goal is not to present a highly technical review of the way polls are conducted. However, if one wants to understand the functions of polling, one must have a basic comprehension of issues such as:
- A.Non-attitudes
- B.Wording and Context of Questions
- C.Data Collection
- D.Sampling.
A Short History of Polling
Manifestations of public opinion research are ubiquitous today. The content we are served when we access our personal social media platforms; the recommendations we get from Netflix, Google, or Amazon; the issues that dominate the political discourse and those that are ignored; the design and execution of communication or corporate campaigns; who runs for political office and who does not are a few examples of how much we rely on this practice. The fascination we have with polling is tied to our essential need to know what others think. Whether it is a politician whose reelection depends on understanding vote intent; a business person whose livelihood rests on offering products or services which will meet the needs of customers; or a social activist needing to mobilize people to a cause, success, or failure relies on a comprehension of what matters to people and how they make up their minds. How we got here is the culmination of a historical process dating back centuries.
The general concept of public opinion was addressed in the works of Plato and Aristotle as well as subsequently by the Romans (Bardes and Oldendick 2017: 3). However, Jean-Jacques Rousseau was the first to use it in the modern sense in 1744 (Bardes and Oldendick 2017: 3; Erikson and Tedin 2016: 1). For him, public opinion evoked the “customs and manners of all members of society as opposed to some elite” (Erikson and Tedin 2016: 1). As years passed, the idea emerged from the Enlightenment that there was a mass public able to exercise its sovereignty and this was key in shaping the role of the public in democratic societies. (Bardes and Oldendick 2017: 3). Simultaneously, efforts were made to move beyond simply discussing public opinion to trying to measure it.
We can find several early attempts by governments around the world to gauge the public’s views on issues. As far back as 1745, the controller general in France undertook a study to gather socio-demographic and economic data in order to examine general attitudes toward a planned increase in land taxes (Blondiaux 1998: 52). About twenty years later, the French minister of the interior Jacques Necker instructed regional commissioners to administer questionnaires to collect information about l’esprit public (the public mood) of the nation (Turcotte 2010: 200). The results of this study were published in the Compte rendu sur l’état général des finances de la France (Blondiaux 1998: 39) and by 1802, the French government was systematically measuring the mood of the French public (ibid.: 53–54). Such efforts were not restricted to France. Elisabeth Noelle-Neumann traces the origins of polling in Germany to social inquiries of the nineteenth and early twentieth century. Examples of such studies include a study sent to 15,000 landowners about the State and the Farm Worker in 1874–1875; “The State of Factory Workers and Apprentices” comprised of questionnaires sent to 7000 factory proprietors, workers, and apprentices in 1875; and “The Morale in the Countryside” which questioned 14,000 Protestants pastors in 1895 (Worcester 1983: 44). In England, Charles Booth, among others, developed methods for collecting and analyzing data that became known as the “English Social Survey” in the early 1800s while in 1915, the US Department of Agriculture began conducting surveys on the quality of life in rural America (Bardes and Oldendick 2017: 19). All those disparate attempts to measure public opinion—as well as the straw polls mentioned at the opening of this chapter—were valiant efforts but lacked in the rigor that would come to define public opinion research.
The road to scientific public opinion measurement proved tortuous. While early attempts to measure public opinion produced questionable results, “intellectual advancements in statistics and the science of sampling were laying the foundations to the emergence of scientific polling industry” (Turcotte 2010: 206). The roots of scientific polling are found in the debates held at the International Institute of Statistics between 1895 and 1903. The Norwegian statistician A. N. Kiaer first failed to convince the Institute of the validity of his method of representative sampling, and this on three separate occasions in 1895, 1897, and 1901. In fact, after his first attempt, opponents of this new approach “warned that this was a dangerous doctrine and argued that a sample could never replace full coverage” (Converse 2009: 41). Finally in 1903, he was more successful and the Institute accepted his methodological assertions (Lachapelle 1991: 4). Around the same time, a statistician-economist at the London School of Economics, Arthur L. Bowley, made robust demonstrations of the validity of representative sampling solidly anchored in probability theory and the central limit theorem (Converse 2009: 42). Another important development occurred in 1925 when the Danish statistician Adolph Jensen presented—to the same Institute—his findings about what would become known as quota sampling (Blondiaux 1998: 171). Finally, R. A. Fisher introduced the principle of randomization (Converse 2009: 44) which became central to the practice of polling.
The details about how sampling entered the political realm and created a new industry are generally well-known. An unknown University professor named George Gallup used his doctoral thesis on sampling techniques to help his mother-in-law become the first woman to hold the position of secretary of state in Iowa in 1932. Three years later, Gallup, as well as Archibald Crossley and Elmo Roper began conducting polls on a regular basis. Despite Gallup’s subsequent notoriety, it was Elmo Roper who was the first of the media pollsters. In July 1935, he released the findings of a study comprised of 3000 American adults about their attitudes toward a range of current affairs issues in Fortune magazine. Roper would repeat this exercise on a quarterly basis and discuss his findings under the rubric entitled “The Fortune Survey” (Blondiaux 1998: 158). A few months later, in October 1935, George Gallup started releasing his own poll results in over sixty subscribing American newspapers in a weekly column entitled “America Speaks” (ibid.). Archibald Crossley followed suit and developed a regular poll entitled “The Crossley Political Poll” for the newspapers part of the Hearst Corporation (Micheau 2018: 29). The first big test of this new way of measuring public opinion came during the 1936 US Presidential election. Gallup publicly challenged the venerable Literary Digest and predicted that the Digest would be unable to accurately predict the outcome of the Presidential election because of inherent flaws in its methodology. As it turned out, Gallup’s own prediction in that election was off by 7 percentage points (Warren 2001: 87), but unlike The Digest, he correctly predicted Roosevelt’s victory and with that, the polling industry was born.
While Roper may have been first, Gallup quickly became the most prominent of the early pollsters and almost singlehandedly expanded polling internationally. Within five years, Gallup opened affiliates in Great Britain (1937), France (1938), and Australia (1941) (Micheau...