To make inferences about social relationships among people, many sociological studies utilize data obtained from a few randomly selected persons. It is also possible, however, to discern and understand social relationships by carefully selecting and analyzing data from a nonrandom sample. Lazarsfeld and Berelson imply that many conclusions obtained from limited studies of voting patterns in Sandusky, Ohio, and Elmira, New York, are true to national voting patterns.' They feel that inferences about voting patterns in the national population are not misleading, because relationships between dependent and independent variables (vote and class or religion) occur all over the United States. Their finding in Sandusky that Catholics are more likely to vote for Democrats and Protestants for Republicans has held up when tested by others using random samples of the U.S. population.2 However, since the proportion voting Democratic in Sandusky and Elmira is not valid for the national population, they cannot predict who will win. The strength of their analysis rests upon the assumption that social factors associated with voting choice (or drug use, in this study) are very similar in widely different parts of the nation. If widely divergent findings emerge in two different studies, they must be empirically demonstrated and explained.

In a manner parallel to Lazarsfeld and Berelson, inferences can be made about social factors associated with college student drug use. The focus will be on the relationship between one or more independent variables and drug use. Other comparable studies will be shown to report similar relationships between variables. However, the incidence of drug use reported in this study is higher than that in the national college population. Nevertheless, the differences are not great. This study focuses upon why college students use drugs and what the consequences of use are, not on how many college students use drugs.

A. STUDY DESIGN

This book reports on the use of various drugs by students at twenty-one different colleges and universities. Most of the colleges and universities are located in the New York metropolitan area, but a few are located outside it. This highly urbanized area has a long history of drug use, although not among its college population. Since World War II, the illicit use of drugs has mainly been confined to heroin use among black and Puerto Rican populations and to some marihuana use among musicians and bohemians in the Greenwich Village area.

New York colleges were among the first affected by trends toward drug use. Samuel Pearlman noted in spring 1965 that 6% of Brooklyn College seniors had tried some drugs illicitly.3 The drug use of these seniors was about two years ahead of that in the first national study of college students; in spring 1967, Gallup reported that 5% of the nation's college students had tried marihuana. Thus, New York students probably reflect emerging trends in drug use among college students that are likely to be followed by the rest of the country. Hence, the New York metropolitan area is an important location for a study of drug use among urbanized college students.

The institutions participating in this study will not be directly identified. In order to obtain the cooperation of an institution, the researcher promised officials at each college that their school would not be identified. In addition, assurances of anonymity were clearly stated on the cover letter attached to all questionnaires: "In the final report, there will be no identification of the classes, individual teachers, administrations, or the institutions (colleges, organizations, etc.) which take part in this study." Thus, all colleges and universities, and even the department from which the students were sampled, have been given pseudonyms.

The sampling design was influenced by three main considerations: (1) A need to obtain a large sample of persons exhibiting heavy drug use, a relatively rare form of behavior, to permit the empirical testing of various hypotheses that attempt to explain drug use. It was desirable to sample college classes where a high proportion of heavy drug users was apt to be found. (2) A need to protect the anonymity of the respondents. It was felt that only assurances of complete anonymity would enable heavy drug users to honestly report their heavy use. Complete anonymity would also make it impossible for police or college officials to trace respondents. (3) A need to reduce expenditures of time and money.

With meager resources a large sample of respondents had to be obtained, the data had to be coded, and then it had to be punched on cards. This restriction dictated the use of a precoded questionnaire to obtain the data and the use of mechanized card punching directly from mark-sense answer sheets. Despite restricted resources, 3500 usable questionnaires were obtained between February 4 and May 6, 1970. The data from these precoded questionnaires, recorded by the respondent on special mark-sense answer sheets, was then converted to punched cards with the aid of an Optical Scanning machine.

Since regular drug use is a reasonably rare phenonema and an increased proportion of such users in the sample was needed, the sample was taken from social contexts in which a higher than average proportion of drug users was known to exist. This was done by sampling from a department in each college that, on the basis of other studies of college drug use, was likely to have a higher than average proportion of drug users and by requesting students who had completed the questionnaire in class to ask heavy drug-using friends to complete the questionnaire and return it to the researcher. Although this sampling design should result in the overrepresentation of regular drug users, many students in a typical classroom will not be drug users. Thus, our sample was designed to increase the proportion of drug users so that important comparisons between regular and less-regular users or nonusers could be made. A random sample of college students in the New York metropolitan area should show a higher proportion of those who did not use drugs and a lower proportion of regular drug users than was the case in the present sample.

The basic instrument for measuring the use of drugs was a precoded questionnaire in which the respondent had to choose from the alternatives given. The questionnaire was similar to the Scholastic Aptitude Test (SAT), which most of the students had taken for admission to college. Each question had ten or less possible answers. The students selected the best answer(s) to the question and marked their response on a special answer sheet. (The Questionnaire and a typical completed answer sheet are provided in Appendix C.)

A precoded questionnaire has several advantages: (1) It protects the respondent's anonymity since all answer sheets look alike. (2) The questionnaire can be used several times, reducing expenses of duplication. (3) Very little recoding of vague answers is needed. (4) The expensive and time consuming task of coding and keypunching is reduced to a minimum. (5) The optical scanning device is extremely accurate (99.5%) in converting the student's answers into punched data; the random errors that ordinarily occur in the coding and keypunching process are drastically reduced.

There are also disadvantages to a precoded questionnaire: (1) Respondents have a severely limited choice of answers; the qualitative richness of open-ended questions is drastically reduced. (2) Students who do not read quickly or who have difficulty with written instructions will have difficulty reading the questionnaire and marking the appropriate response on a separate sheet. (3) Students may easily mark the wrong response or mark the wrong question. The advantages of this technique, however, are probably greater than the shortcomings. The precoded questionnaire is most efficient when used to collect a large amount of comparable data in a short period of time. The researcher may proceed with data analysis more rapidly.

B. SELECTION OF THE SAMPLE

Colleges in America vary greatly in student-body composition. Different schools attract different kinds of students and, hence, have widely different rates of drug use.4 It was therefore thought that the colleges selected should represent the range of variation that exists in institutional types. Two characteristics were used in selecting the colleges for the sample, the quality of students attending the college and the type of institutional control, private, religious, or public.

Obtaining an objective measure of the academic caliber of students in different colleges seems fairly difficult. However, most colleges in the New York area require and compile SAT scores for incoming freshmen. The average SAT scores for each college are presented in Singletary.' These scores permit the classification of colleges according to the quality of their incoming freshmen. This is probably a fairly reliable measure of the quality of the college as well.6

Singletary reports both the average verbal and average mathematical SAT scores. Although his statistics are for the freshman class of 1967, there has probably been little change in the relative positions of the colleges on mean SAT scores for incoming freshmen. The verbal SAT scores were selected as the best indicator of quality, mainly because the mathematical SAT scores are strongly affected by the sex ratio of the college. The cutting points were arbitrarily chosen, but they probably reflect differences in quality among the colleges in the sample:

The second major way in which colleges were classified was according to the type of institutional control. It is assumed that the agencies controlling colleges strongly affect the individual and social selection processes that bring students to a particular college. There are important Socio-Economic Status (SES) differences between students attending private and public secular colleges. Students at religious colleges are recruited disproportionately from their own religious denominations, and they are probably the most devout of those in that religion. Further, pretesting of the questionnaire indicated some important differences in drug use between religious groups.

Table 1 contains a list of colleges, none of which were in the sample, by quality and type of control so that the reader can obtain some idea of the kinds of colleges sampled in this study. Table 2 contains the list of colleges actually sampled, but only a pseudonym for the college is supplied. It also contains information about the quality, type of control, sample size, approximate proportion of the total daytime student population sampled, and the percentage who have ever used marihuana or hashish. It was possible to obtain between 2% and 8% of the total student population at each school, which constituted a fairly adequate sample of all schools but two (discussed in Appendix B). It is interesting to note that the colleges having the highest (Old) and lowest (St. Luke's) levels of cannabis use are both in New York City.

With the very limited resources available for the study, the decision was made to limit the sample to one major department in which many drug users might be found. The literature in the field of drug use indicates that persons majoring in the social sciences and humanities (or liberal arts) have higher rates of drug use than those majoring in other fields,' although the evidence is not overwhelming. It was decided to randomly sample students enrolled in courses offered by one major social science department, henceforth referred to by its pseudonym, the Ethics Department. It was expected that gaining access to classes in other departments would be more difficult because they would be less willing to devote instructional time to a drug survey; this proved to be the case. Furthermore, the Ethics Department was very cooperative, the refusal rate of instructors was very low, and 75% of all Ethics instructors contacted agreed to let their class complete the questionnaire during a regular class period suggested by the researcher.

The Ethics Department provided a fairly broad spectrum of students. The introductory course in Ethics is taken by many who do not plan to major in Ethics; however, many major in other social sciences or the humanities. Since most colleges require a minimum number of social science credits for graduation, there were some students from every major. The Ethics majors did prove to have a high rate of illicit drug use, somewhat higher than majors from other social science departments and considerably higher than science, business, and education majors. Thus, the initial assumption of a high rate of drug use in the Ethics department proved to be correct. Random selection of Ethics Department classes should provide about the same kind of students in the twenty-one different colleges. It is assumed that students attending classes in the Ethics Department at college A are more like the student body of college A than they are like students in the Ethics classes of college B, where college A and B are very dissimilar.

The most efficient way to collect a large number of cases is to give the questionnaire during a regular class hour. Classroom administration has the advantage that most students present will complete the questionnaire. Only a few students (about 2%) will walk out or refuse to participate in the study. The major disadvantage is that many students (10-15%) will be absent on a particular day, with absenteeism being especially high toward the end of the semester.

After a college had been selected, the schedule of classes and room assignments for spring 1970 were obtained from the college registrar or Ethics Department secretary. All classes (and sections) listed by the Ethics Department were considered part of the sampling frame. These classes were divided into two strata: introductory classes, which can be taken without any prerequisites in the Ethics Department, and advanced classes, which usually require the student to have taken a previous course in the Ethics Department. This division was considered important because introductory classes have large numbers of students from other majors. They are therefore more representative of the student body as a whole than are advanced Ethics classes. However, a large sample of Ethics majors was desirable; pretest evidence suggested that a higher proportion of heavy drug users would be found in the advanced classes. Hence, about an equal number of introductory sections and advanced classes were sampled at each college, but more introductory students were actually obtained than advanced students (there were about three introductory students for every two advanced students sampled) because of differences in class size.

All introductory Ethics sections were numbered from 1 to N in the order that they were listed in the catalogue, where N is the number of introductory sections in the Ethics Department. Then, using a table of random numbers,8 the first random digit equal to or less than N was the first section selected as part of the sample. Then, the next random number less than or equal to N was the next section chosen, and so on, until enough classes were sampled to provide 75-125 students attending introductory classes at the college. The same procedure was followed to select the advanced Ethics classes; 3-6 classes were usually sampled.

At the same time that the sample was being selected, the researcher made contact with the chairmen of the Ethics Department in order to obtain permission to contact instructors. Frequently the necessary permission to contact instructors was gained by a telephone call or a short personal visit to the department chairman in which the purpose and extent of the study was explained. At 50% of the colleges this contact was sufficient to gain the cooperation of the whole department. Other chairmen requested a copy of the questionpaire and/or a letter from the researcher's Ph.D. adviser before making a decision. A week or two later the chairman would either give permission to contact instructors or refer the researcher to the dean of students for his consent.

Once the consent of the chairman or dean was obtained, a form letter was sent to the instructor of each class in the sample. In this letter the researcher indicated the class he desired to sample and suggested a convenient date and time to give the questionnaire. An attached postcard allowed the instructor to indicate whether his class would cooperate or not. This was a most effective way of gaining cooperation, and little difficulty was encountered in gaining the consent of the teachers until midterm-exam time.

Once the cooperation of the department chairman and instructor had been obtained, a date and time convenient for both researcher and instructor (usually the time indicated in the letter sent to the instructor) was agreed upon. The researcher encouraged the teacher to inform the class about the study and let the students vote on whether or not they wished to participate. The instructor was requested not to inform the class of the exact date that the questionnaire was to be administered. This way the students would not feel that something had been forced on them by the instructor, since by voting they had chosen to participate in the study voluntarily.

On the date agreed upon, the researcher or an assistant attended the class, briefly explained the purpose of the study, handed out the questionnaires, answer sheets, and pencils, and told the students to begin. The students followed the written directions on the questionnaire. Since the questionnaire was complicated in some parts and very long (taking most students 40-60 minutes to complete), the researcher's presence was generally necessary to answer questions and to urge slow students to skip certain questions and complete other questions if they were obviously not going to finish in the class hour.

Two additional, nonrandom groups were included in the sampling design. First, the researcher tried to obtain the cooperation of some instructors and classes outside of the Ethics Department. It was hoped that this would give some indication of how biased the sample from the Ethics Department was. Unfortunately, considerable effort provided little cooperation from non-Ethics instructors. Second, in an attempt to obtain an even greater number of regular drug-using respondents, all students in class were requested at the end of the hour to give a copy of the questionnaire, an answer sheet, and a stamped self-addressed return-envelope to a friend who had used more than six times any of the following drugs: hallucinogens, methedrine, cocaine, or heroin. About 10% of the students in most classes took questionnaires home to friends. From such friends 165 usable questionnaires were returned to the researcher.

C. VARIATIONS IN DRUG USE BY MAJOR SAMPLING FACTORS

Graph 2.1 shows the proportion of students ever using marihuana by the major variables on which the sample was stratified. One finds that cannabis use increases with the quality of the college. But schools of the lowest quality have a fairly high rate of cannabis use. Likewise, the type of institutional control influences cannabis use, but differences are slight. Catholic colleges have a significantly lower incidence of cannabis use. The expectation that drug use among students in Ethics classes is greater than among students in non-Ethics classes and that students in advanced Ethics classes have higher levels of cannabis use than those in introductory classes is verified. But cannabis use is highest among the friends who mailed in the questionnaires. This group included a few who did not use drugs, despite directions to give the questionnaire only to regular drug-using friends. Thus, our expectations as to variation in drug use among subgroups specifically selected to be part of the sample are verified.

The sample actually obtained departs in some ways from the original sampling design but probably not enough to substantially alter any of the fundamental relationships between various independent variables and drug use. For example, 70% of all colleges that were ever part of the sampling frame participated in the study; 77% of the randomly selected Ethics classes and 97% of the students attending class on the day that the questionnaire was administered cooperated in this study. Thus, all indicators of participation in this study show that a relatively adequate sample was obtained.

Rather than present the technical problems of obtaining the sample in the main body of the book, the interested reader is referred to Appendix B, in which several data-collection problems, such as obtaining cooperation from colleges, classroom instructors, and students, are discussed. Difficulties that arose in obtaining cooperation from non-Ethics departments and from heavy drug-using friends are dealt with, as well as problems with mark-sense machines.

REFERENCES

1. Paul Lazarsfeld, Bernard Berelson, and Hazel Gaudet, The People's Choice, 3rd ed., New York: Columbia University Press, 1968. Berelson, Lazarsfeld, and William McPhee, Voting, Chicago: University of Chicago Press, 1954, p. 297.
2. Ibid., pp. 331-347. Angus Cambell et al., Elections and the Political Order, New York: John Wiley and Sons, 1966, esp. chap. 6 by Philip E. Converse, "Religion and Politics: The 1960 Election," pp. 96-124. Similar religious differences emerge in other Anglo-American countries: Robert R. Alford, Party and Society, Chicago: Rand-McNally, 1963.
3. Samuel Pearlman, "Drug Use and Experience in an Urban College Population," American Journal of Orthopsychiatry, 38 (April 1968), 503-514.
4. Richard Blum et al., Students and Drugs, San Francisco: Jossey-Bass, 1969, p. 41. Edwin S. Robbins et al., "College Student Drug Use," American Journal of Psychiatry, 126 no. 12 (1970), 1747.
5. Otis Singletary, American Universities and Colleges, 10th ed., Washington, D.C.: American Council on Education, 1969.
6. Paul F. Lazarsfeld and W. Thielens, Jr., The Academic Mind, New York: Free Press, 1958, indicate that several different measures of college quality are relatively interchangeable.
7. Blum, Ref. 4, p. 65.
8. Samuel B. Richmond, Statistical Analysis, 2nd ed., New York: Ronald Press, 1964, pp. 595-596.