It seems obvious that in a democracy, the government should be guided by public opinion: after all, democracy means rule by the people. It is difficult, however, to say exactly what public opinion is: âto speak with precision of public opinion is a task not unlike coming to grips with the Holy Ghostâ (Key 1961, p. 8). There are polls and surveys on many different topics, but these attempts to measure public opinion raise a number of questions. One is simple accuracy: can a survey given to 1,000 people tell us about the opinions of the whole nation? Others involve the interpretation of answers. Sometimes different surveys seem to point to different conclusions. In other cases, the results seem clear, but the underlying issues are complicated ones about which most people are not well informed. For example, recent surveys in the United States show widespread support for raising the minimum wage to $15.00 an hour, but many economists think that this would lead to a substantial increase in unemployment, which people presumably do not want. Suppose that these economists are correct: should we then conclude that the public does not really favor an increase, because it does not understand what effects it would have? Or should we say that the public thinks that an increase in the wages of low-wage workers is more important than any effect on employment, or that it favors some increase in the minimum wage, but is not committed to an exact number?
This chapter will address questions of measuring and interpreting public opinion. Subsequent chapters consider the relations between society and public opinion. There are often differences of average opinion among groups, such as social classes, racial and ethnic groups, or residents of cities and rural areas. Chapter 2 considers the reasons that such differences appear, persist, and change. Chapter 3 discusses the organization of opinions, particularly the distinction between left and right (liberal and conservative). Chapters 4 and 5 consider change in opinions: Chapter 4 focuses on the period for which survey data are available â from the 1930s to the present â while Chapter 5 takes a longer view and asks whether there are trends that extend over centuries. Finally, Chapter 6 considers how public attitudes toward government are changing, and what these changes mean for the future of democracy.
The Rise of Public Opinion
The expression âpublic opinionâ first appeared in the late 1700s, and soon came into wide use. During the nineteenth century, many observers spoke of the power of public opinion. In 1823, Lord John Russell, a Member of Parliament and future Prime Minister of Great Britain, wrote: âit is the fashion to point out the increased and increasing influence of public opinionâ (Russell 1823, p. 429). A few years later, a popular American college textbook called public opinion âthe sense and sentiment of the community, necessarily irresistible, showing its sovereign power everywhereâ (Lieber 1839, p. 253). At about the same time, Alexis de Tocqueville (1850 [1969], p. 124) wrote that public opinion was the âdirecting powerâ in both the United States and France, despite the differences in their forms of government: âin America it works through elections and decrees, in France by revolutions.â
The appearance and growth of âpublic opinionâ resulted from a change in the relationship between government and society. Plamenatz (1975, p. 345) defines public opinion as âopinions about the government and its policies current in circles outside the [government] hierarchy and yet close enough to it to acquire such opinions and to bring them to bear on it.â For most of history, only a small elite group could regularly participate in government; most people had no way to even become aware of what policies or actions the government was considering. Ordinary people sometimes tried to influence the government through collective protests, but these usually involved objections to the conduct of local officials or landowners rather than attempts to change government policy (Tilly 1983). This situation started to change with the Industrial Revolution, as more people became aware of the government and acquired more means to influence it. The sources of this change included the spread of literacy, the appearance of newspapers and magazines, the growth of cities, and the expansion of the âmiddle classâ â people with enough knowledge and leisure time to pay attention to public affairs. These developments meant that news could spread more quickly and be discussed more widely, so that people could form opinions about government policy and organize to influence it.
At first, public opinion was often understood to mean middle-class opinion, but the range of the âpublicâ expanded as time went on. The weakening of property restrictions on voting and the general adoption of the secret ballot in the late nineteenth century were important parts of this process, since voting gave ordinary people an easy and inexpensive way to influence the government (Rokkan 1961). At the same time, rising educational levels and the growth of the mass media made it easier for people to be informed about public affairs. Today, the âpublicâ is generally understood to include the entire adult population, and people have become accustomed to offering opinions on all kinds of topics.
Although nineteenth-century observers agreed that public opinion was important, they found it difficult be sure of what that opinion was on any given question. In an election, voters merely choose a party or a person: they do not get to vote on specific policy proposals. Moreover, candidates often give different messages to different audiences or use ambiguous language that can be interpreted in a variety of ways. As a result, we know that voters preferred something about the winner, but do not know exactly what that was. People can express opinions more precisely by letters and petitions to government officials, or by public rallies and demonstrations, but only a small fraction of the public engages in such actions, and it is possible that the opinions of the people who do not are very different from the opinions of those who do. Moreover, the opinions of the people who do not participate cannot safely be ignored: they might resist a policy after it is enacted or vote against the government at the next election. The positions taken by organizations and the opinions expressed in the media can also be taken as indicators of public opinion, but these have a similar limitation: the members of an organization might not share the views of the leaders, and the readers of a publication might not share the views of the writers.
Public Opinion Surveys
In the 1930s, a new form of measurement appeared which transformed the study of public opinion: the survey. The basic procedure of an opinion survey is to ask a standard set of questions to a group of people who are supposed to represent the public (the âsampleâ). Surveys are conducted in the same way that elections are: participants (also known as ârespondentsâ) answer the questions in private and are assured that their individual answers will not be disclosed. Interviewers are instructed to be neutral â to simply record answers without raising objections or expressing their own opinions. Usually the participants choose their answer from a standard list, rather than responding in their own words. In effect, every survey question can be regarded as a small-scale referendum conducted by secret ballot. George Gallup, who founded the first survey organization, saw surveys as a way to improve the operation of democracy by giving political leaders a more accurate and detailed picture of public opinion (Gallup 1938).
In addition to giving information on opinions in the public as a whole, surveys can include background questions on characteristics such as race, gender, and educational level, making it possible to distinguish the opinions of different kinds of people. Moreover, most surveys ask for opinions on a number of topics, so it is also possible to examine the relationships among different opinions, or between opinions and voting choices.
The first surveys were conducted by commercial organizations and focused on predicting elections and measuring opinion on issues of the day (surveys of this kind are often known as âpollsâ). If an issue remained prominent, the polls sometimes repeated questions that they had previously used. For example, the Gallup Poll asked âare you in favor of labor unions?â in July 1936, and asked the same question again in 1937, twice in 1938, three times in 1939, and twice in 1940. When questions are repeated, it is possible to examine changes in opinion, both in the general public and in specific groups. For example, in June 1937, 70 percent said that they were in favor of labor unions and 22 percent that they were not; in October 1938, 58 percent were in favor and 28 percent were not.
Academic researchers soon began to conduct surveys, and sometimes made systematic efforts to repeat questions on a regular basis. The American National Election Studies, which began in 1948, focused on voting and political opinions, while the General Social Survey began in 1972 and covered a wide range of topics. At the same time, other polls and surveys continue to repeat questions more or less frequently. As a result, there is now a record of change in public opinion, which on some topics extends over a period of more than eighty years.
Surveys were soon adopted in several other nations, including Canada, France, Australia, and Great Britain. National survey organizations sometimes agreed to include the same question, or sometimes several questions, in their polls, making it possible to compare opinion across nations. Academic researchers followed with more systematic efforts to develop âcomparative surveys,â in which an entire survey was translated into the local language and administered in different nations. Because of the expense and organizational effort necessary to carry out comparative surveys, only a few were conducted until the 1970s. Since that time, however, they have become more common, and some of them are part of continuing series that repeat questions over time. The Eurobarometer, which began in 1974, includes nations in the European Union. The World Values Survey began in 1981, when it included ten nations; the seventh wave, conducted in 2017â2020, will include about eighty nations from all parts of the world. Since 1985, the International Social Survey Programme has conducted an annual survey focusing on a particular topic: examples include the role of government in 2016, social networks in 2017, and religion in 2018.
The accumulation of survey data has increased the range of research that is possible: analysts are able to compare not only different kinds of people, but also different places or times, or all of these levels at once. For example, Brooks, Nieuwbeerta, and Manza (2006) used data from 112 election surveys to compare the effects of gender, class, and religion on voting choices in six nations between 1964 and 1998.
Another important recent development has been the increased use of survey experiments, in which different respondents in a survey are randomly selected to receive different information or different forms of a question. Simple experiments have been used since the beginnings of survey research but when surveys are given over the internet, it is possible to have more complex designs and use a wide variety of cues â for example, asking respondents to read a passage or watch a video before they answer a question.
Accuracy of Opinion Surveys
Sampling
A basic problem in survey research is how to obtain a representative sample â a sample that is like the population in all respects, except that it includes a smaller number of people. At first, most surveys sought to achieve this goal by âquota samples.â In this method, interviewers were given quotas for certain characteristics that were thought be important for opinion, such as gender, race, and age, and were otherwise left free to choose respondents in any way they saw fit. For example, an interviewer might be instructed to obtain twenty interviews, which would include ten men and ten women, seventeen whites and three blacks, four people aged 18 to 29, twelve people aged 30 to 64, and four people aged 65 and above (see Berinsky 2006 for a more detailed description of the procedures used in early surveys). Although this method guaranteed the sample would be representative in terms of the characteristics for which there were quotas, it did not necessarily make it representative in other respects. For example, if interviewers obtained their samples by approaching people in public places, then people who rarely left their homes or who worked unusual hours would be underrepresented.
An alternative way of obtaining a representative sample is a random (probability) sample, in which every person is assigned a definite chance of being chosen for the sample, and the decision of whether to include them is made at random â in effect, by a lottery. There is no guarantee that a particular random sample will be exactly representative, but it is likely to be close to the population. Suppose that a random sample of 1,000 people is drawn from a population in which 50 percent are women. It is likely (about an 80% chance) that between 48 and 52 percent of the people in the sample will be women, and almost certain (about a 99.9% chance) that between 45 and 55 percent will be women.
Random samples have three major advantages over quota samples. The first is that they will be approximately representative in terms of all characteristics, not just the ones that are included in the quotas. The second is that as a random sample becomes larger, the distribution of characteristics in the sample tends to come closer to the distribution in the population. This means that the accuracy of the sample estimates can be increased by increasing the size of the sample. For example, suppose that the population is 50 percent female. The chance that the sample will be within two percent of that figure â that is, between 48 and 52 percent female â rises from about 80 percent in a random sample of 1,000 to about 97 percent in a sample of 3,000 and 99.5 percent in a sample of 5,000. In contrast, with other methods of sampling, increasing the size of the sample will not necessarily bring it closer to the population. The third advantage of a random sample is related to the second: given the size of the sample, it is possible to calculate a âmargin of errorâ â that is, to estimate how much difference there might be between the sample and the population.
In the United States, most surveys shifted from quota samples to random samples after about 1950. This change was facilitated by the growth of telephone ownership, which made it easier to obtain a random sample. If everyone has a telephone, then a random sample of the public can be obtained by simply dialing randomly generated telephone numbers. Quota sampling lasted longer in many other countries, but random sampling has come to be the standard method. Experience shows that quota sampling often was reasonably effective, and statistical adjustments can be applied to produce a closer match to a representative sample (Berinsky, Powell, Shickler, and Yohai 2011). Nevertheless, random sampling is preferred because it has a firm theoretical foundation â we know that a random sample is likely to be accurate within limits determined by the size of the sample.
Non-Response
In principle, random sampling provides a definitive solution to the problem of choosing a representative sample. In practice, however, researchers do not have complete control over the sample. Survey researchers can attempt to contact a random sample of the public, but some of the people in that sample will not be at home, and others will refuse to participate. If the people who do not participate are different from those who do â ...