the preferences, beliefs, and attitudes of given samples of people.
Ask Yourself
Has the increasing legalization of marijuana throughout the United States altered the data on marijuana use among high school seniors? Why or why not? How might any change affect data on the use of other drugs?
For many years, the Institute for Social Research at the University of Michigan has conducted a descriptive survey of high school seniors in the United States. One of the subjects has been marijuana use. As you can see in Figure 2.3, the prevalence of marijuana use among high school seniors has risen and fallen, as if in waves. Marijuana use in this group peaked in 1979 (with more than half of students admitting use of the drug), reached a low of 22 percent in 1992, and has generally risen since then, although it has never again approached the 1979 level. In 2017, 37 percent of twelfth graders reported having used marijuana in the previous 12 months.
Figure 2.3 Marijuana Use Among U.S. High School Seniors, 1976–2017
Source: Data from Lloyd D. Johnston, Patrick O’Malley, Richard A. Miech, Jerald G. Bachman, and John E. Schulenberg. Monitoring the Future: National Survey Results on Drug Use, 1975–2017: Overview, Key Findings on Adolescent Drug Use, Table 6 (Ann Arbor: Institute for Social Research, University of Michigan, 2017).
The data in Figure 2.3 are derived from descriptive surveys, but what if we wanted to explain, and not just statistically describe, changes in marijuana use among high school seniors? To get at this, we would need to do an explanatory survey, which seeks to uncover potential causes of, in this case, changes in marijuana use (e.g., the legalization of marijuana in states such as Colorado, California, and Maine [Monte, Zane, and Heard 2015]). For example, having discovered variations in marijuana use by high school students over the years, we might hypothesize that the variation is linked to students’ (and perhaps the general public’s) changing perceptions about the riskiness of marijuana use. Specifically, we might hypothesize that as students (and the public) increasingly come to see marijuana as less risky, marijuana use among students will go up. In this case, we would use the survey to learn more about respondents’ attitudes toward and beliefs about the riskiness of marijuana use and not simply measure student use of marijuana.
Sampling
It is almost never possible to survey an entire population, such as all Americans, all students at your college or university, or even all sorority members at that university. Thus, survey researchers usually need to construct a sample, or a representative portion of the overall population. The more careful the researcher is in avoiding biases in selecting the sample, the more likely the findings are to be representative of the whole group.
The most common way to avoid bias is to create a random sample, a sample in which every member of the group has an equal chance of being included. One way of obtaining a random sample is by using a list—for example, a list of the names of all the professors at your university. A coin is tossed for each name on the list, and those professors for whom the toss results in heads are included in the sample. More typical and efficient is the use of random number tables, found in most statistics textbooks, to select those in the sample (Kirk 2007). In our example, each professor is assigned a number, and those whose numbers come up in the random number table are included in the sample. More recently, use is being made of computer-generated random numbers.
Other sampling techniques are used in survey research as well. For example, the researcher might create a stratified sample in which a larger group is divided into a series of subgroups (e.g., assistant, associate, and full professors) and then random samples are taken within each of these groups. This ensures representation from each group in the final sample, something that might not occur if one simply does a random sample of the larger group.
Random and stratified sampling are the safest ways of drawing accurate conclusions about a population as a whole. However, there is an element of chance in all sampling, especially random sampling, with the result that findings can vary from one sample to another. Even though sampling is the safest way to reach conclusions about a population, errors are possible. Random and stratified sampling are depicted in Figure 2.4.
Sometimes researchers use convenience samples, which avoid systematic sampling and simply include those who are conveniently available to participate in a research project. An example of a convenience sample might involve researchers passing out surveys to the students in their classes (Lunneborg 2007). These nonrandom samples are rarely ever representative of the larger population whose opinions the researcher is interested in knowing. Nonrandom samples therefore may create a substantial bias in researchers’ results (Popham and Sirotnik 1973). Many surveys that pop up on the internet are suspect because the respondents are the people who happened to be at a certain website (which is likely to reflect their interests) and who felt strongly enough about the topic of the survey to answer the questions.
Research using convenience samples is usually only exploratory. It is almost impossible to draw any definitive conclusions from such research. There are, however, some cases (e.g., studying a group in which many members are reluctant to be studied) in which convenience sampling is not only justified but also necessary and useful. Convenience sampling also sometimes leads to larger, more scientific projects that rely on random or stratified samples.
Experiments
Some sociologists perform experiments (Jackson and Cox 2013). An experiment involves the manipulation of one or more characteristics in order to examine the effect of that manipulation (Kirk 2007).
A study by Devah Pager (2009) is a good example of a sociological experiment. Pager was interested in how the background of a job applicant affects the likelihood of that individual’s being called back for an interview. Pager randomly assigned fake criminal records to pairs of similar young men, one in each pair black and one white. Thus, in each pair, one person had a criminal record and one did not, and one was white and one was not. These young men then sent résumés to companies in Milwaukee, seeking entry-level jobs. One major finding of this experiment was that the young men believed to have criminal records received callbacks less than half as often as did those of the same race believed not to have criminal records.
Figure 2.4 Random Samples and Stratified Samples
Source: Random Samples and Stratified Samples is reprinted with permission of Dan Kernler, Associate Professor of Mathematics, Elgin Community College, Elgin, IL.
In this experiment, we can clearly see the relationship between two important elements of an experiment: independent and dependent variables. In Pager’s experiment, the independent variable, the condition that was manipulated by the researcher, was the job applicant’s combination of race and criminal background. The dependent variable, the characteristic or measurement that resulted from the manipulation, was whether the applicant was called in for an interview.
There are several types of experiments (Walker and Willer 2007):
Laboratory experiments. Laboratory experiments take place in controlled settings. The “laboratory” may be, for example, a classroom or a simulated environment. The setting offers the researcher great control over the selection of the participants as well as the independent variables—the conditions to which the participants are exposed (Lucas, Graif, and Lovaglia 2008).
Natural experiments. Natural experiments are those in which researchers take advantage of a naturally occurring event to study its effect on one or more dependent variables. Such experiments offer the experimenter little or no control over independent variables