Scribd Logo
Upload

Welcome to Scribd!

  • Singleton and Straits - Survey Research

    Uploaded by

    hiandic
    24 views47 pages
    survey

    Original Title

    Singleton and Straits_Survey Research

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    survey

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    24 views47 pages

    Singleton and Straits - Survey Research

    Uploaded by

    hiandic
    survey

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 47
    APPROACHES TO |SOCIAL RESEARCH FOURTH EDITION Royce A. SINGLETON, JR. BRUCE C. STRAITS as Memons of pata eauLeRoN 23. Campetl and Stanley (1963) orginally used the tem “morality” to describe this source af invalidity. However, ation, whichis now commonly used, seems beter cap. ture the intended meaning. 44 However, i the teacher inated the experiment one day becauso of a sharp increase ‘in talking on of turn, the incidence of tis behavior may regres toward a more typical cas com level on the posttest. '. However, this design does not by itself conto fr eects of unintended events that right occur within a teatment group. For a discussion ofthis problem and suggestion for eaing with, see Campbell and Staley (1963:13-14), 6, Notice tow the prinepl of “doing only ne thing at aime” applies in interrting factorial resus. The only diference between the Fistrow and the Second-o¥ groups isthe factor-A manipulation. The to groups ae equivalent on factor B (each row has st equ number of pretested and nonpreested subjects) and shouldbe approximately equivalent on ‘etrancous variables is a result of subject randomization 7, For a more complete discussion, see Camptell (1969) and Cook and Camp! (4976, 1979, 8 “ Survey Research Survey research in its many forms hus become a very common activity in today's ‘world, and most of us have had some experience with it in one form or another. Perhaps you have been stopped on the sttest by a radio news reporter and asked ‘your opinion on some issue of local or national importance. You may have re~ Ssponded toa reader survey found in a popular magazine. Or perhaps you have filled ‘out and returned a brief questionnaire that came with a small appliance or other product you purchased. You or someone else in your houseiold very likely # sponded to the last U.S. Census, which attempted to enumerate and gather con ential information about every person living inthe United States. You have been 4 consumer of survey research if you have read in the newspaper the results of Gallup or Roper public-opinion surveys. You have done a litle “survey research” ‘of your own if you moved to a new community and asked a number of residents about local restaurants or where 1 obtain various services, General Features of Survey Research The survey examples above differ in their degree of formality and in the extent 10 Which they conform to these typical features of professional survey research: 1. A urge numberof respondans are chosen through probebilsy sampling proce resto represent the population of interest. 2, Systematic questionnaire oe inciview procedures are used task prescribed {questions of respondents and record thei answers 3. Answers are numerically coded and analyze [As we now elaborate on these three features through reference to actual studies, we will also describe exceptions tothe general rule for each feature Large-Scale Probability Sampling Professional surveys make use of large samples chosen through scientific sampling procedures to ensure precise estimates of population characterises. National opin {on polls typically number around 1000 respondents, but surveys of national sam- ples can be much larger. Inthe National Educational Longitudinal Study (NELS), for example, the probability sample interviewed in 1988 consisted of 24599 eighth racers from 1052 public and private schools. Participating students were asked 219, 20 Memos oF DATA COLLECTION about a wide range of school, family, and home experiences, and these data were Supplemented by information obtained from paren’ and teachers as wells achieve- ‘ment tes cores and schoo! charatritics sc as teacher and student composition. Beckett Broh (2002) analyzed the NELS data in her study of the impact of ex- tracuricular involvement on academic achievement, Using data fom follow-up surveys of NELS respondents, Thomas Delete and Aviel Kalil (2002) examined the efecto family struture on four developmental outcomes among teenagers: whether they graduate from high school, atend college, smoke or drnk, and initiate sexual activity. As previous researchers had found, youth from nonmarsed familis did ‘not fare as well as youths from marred families om al four outcomes. A major ex- ception to this patter, however, was youths from single-parent muligeneratonal families, who did jus a well as youths from mared families. ‘The NELS was conducted by the National Opinion Rescarch Center on behalf ofthe National Center for Educational Statistics of the U.S. Department of Educa: tion. Such national studies require considerable resources—time, money, and personns!—tha are beyond the capacity of independent researchers or small Search teams. Many surveys, therefore, involve somewhat smaller samples drawn fom state or local populations. The 1979 Detroit Area Study, for example, entailed {interviews with 640 respondents from randomly selected households ina stratified cluster sample ofthe city of Detroit (Young, 1991). The survey focused on attudes towant crime and the criminal justice system. Among the findings was a difference between blacks and whites in the factors that influenced support for the death penalty. Whites, who are far more likely than blacks to support capital punishment, fended to base thei support on perceptions of the causes of criminals’ actions, blacks’ support, however, depended om their degree of trust in the police. Many useful surveys also are conducted with stl smaller samples as well as with noaprobability sampling methods. In 1990 the Worcester Area Project on Ag- ing interviewed 229 people aged 65 or older who were randomly selected from the annual city census of Worcester, Massachuscts (Ainlay, Singleton, and Swigert, 1982), Many other surveys ofthe eldetly, in particular, are conducted with non probability samples. Gay Young and Winifred Dowling (1987), for example, sur veyed 123 members of five El Paso, Texas, chapters of the American Assocation of Retired Persons (AARP), who retumed questionnaires distributed at chapter Iuncheoas. ‘Thus, while large-sale probability samples are the ideal, surveys vary consid- erably in Sample size and sampling design. There are legitimate reasons for doing a small-scale survey, particulary if you havea low budget or some specialized or applied research purpose. Indeed, students can often conduct tee own low-budget research if they have a research problem that can be studied appropriately with a bret questionnaire survey of the home campus or & Ielephone survey of the local dialing area. "Atthe other end ofthe spectrum from surveys of local populations are surveys that draw samples from more than one nation, In comparative or cross-national surveys, equivalent surveys are conducted in different countries. For example, based on interviews with over 25,000 men and women from 22 diferent nations, Haya Stier and Noa Lewis-Epsein (2008) analyzed preferences for the amount of| ‘Survey Research a time spent at work. There is considerable cross-national variation in weekly work ing hours and trends in working hours; for example, during the 1980s and early 1990s working hours increased in the United States and the United Kingdom but declined in Japan and Germany. Stcr and Epsten-Lewis found that preferences for ‘working hours were affected by both individual-level and country-level character- istics, Atte individual level, people with higher education and income and the el- erly prefered to work fewer hours, whereas the opposite was true for those with lower education and income. At the societal level, a highly developed economy and Jw economic inequality were associated with a preference to reduce the time spent at paid work. Units of analysis. While in most surveys the units of analysis are individuals, this isnot always the case. An example ofa survey eating cities as units involved ‘community decisions about whether to fluoridate water supplies to reduce tooth de- cay. A controversial issue in many communities in the 1950s, the proposal to fluo ridate provided social scientists with an excellent opporcunity to study the politieal decision-making process in cities. To learn how cites had dealt with the issue of fluoridation, Robert Crain, Elihu Katz, and Donald Rosenthal (1969) sent ques- ‘ionnaites to three informants in each of 1181 cities: the public health officer, the publisher of the largest city newspaper, and the city clerk. The units of analysis ‘were, again, the cities, One finding was that when te issue was decided by re edu, the odds against fluoridation were five to one. Systematic Procedures: Interviews and Questionnaires Surveys obtain information darough interviews and/or self-administered question- naires, The Detroit Area Study illustrates the use of interviews: the U.S. Census uses a combination of interviews and self-administered questionnaires; the 1969 study by Crain, Katz, and Rosenthal of politcal decision making in cities made use of questionnaires oaly ‘An additional example of the use of questionnaires is Henry Wechsler and as- sociates" (1994) widely cited survey of American college students’ drinking habits ‘The survey's 20-page questionnaire, completed by 17,592 sudents at 140 colleges, asked numerous questions about drinking behavior and other health issues. The ‘most controversial findings concerned binge drinking. Operationally defining binge drinking as the consumption of five or more drinks in @ row for meh and four or _more drinks in a row for women during the two weeks prior t the survey, the T= searchers found that 44 percent ofthe students responding tothe survey were binge rinkers, Binge drinkers, especially those who frequently binged, were far more likely haa noabinge drinkers to experience a variety of alcobol-reated and other health problems, such as engaging in unprotected and unplanned sex, getting in trouble with campus police, damaging property, and getting hurt or injured Unstructured versus structured interviewing. Regardless of whether the sur- vey researcher makes use of interviews only, questionnaires only, or some combi- nation of the two, the procedures tend to be standardized forall respondents ino 22 Semone of asta cotter dr t enhance the reliability of the dats, This epitomizes a structured interview; The objectives are very specific, all questions are written beforehand and asked in, the same order for al respondents, and the interviewer is highly restricted i such matters as the use of introductory and closing remarks, transitions of “bridges” ftom topic to topic, and supplementary questions to gain a more complete response (called “probes"). There are exceptions to this rule, however, as informative and scientifically useful interviewing is sometimes carried out ina less formal or struc~ tured manner. ‘In an unstructured interview, the objectives may be very general, the diseus- sion may be wide ranging, and individual questions will be developed sponta neously inthe course of the interview. The interviewer is free to adapt the inter- view to capitalize on the special knowledge, experience of insights of respondents, ‘An everyday example of an unstructured interview might be a journalist's ints viewing a celebrity to learn more about his or her personal background, interests, and lifestyle. Between the two extremes, the semistructured interview would have specific objectives, but the interviewer would be peemitted some freedom in meet ing them. The scope of the interview would be imited to certain subtopics, and key ‘questions probably would be developed in advance The choige ofa highly structured, semistructured, or unstractured approach de pends on the researcher's objectives, For example, in their study of undocumented migration among rural Mexicans, Douglas Massey and associates (1987) wanted 10 ‘gather quantitatively rigorous data, but ruled out a highly strctared approach be- cause of the sensitive nature of the study and the respondents” limited background (many were poorly educated or illiterate). Their semistructured interview form listed subtopics or data to colleet for each houschold member. The interviewers ‘were given considerable discretion as to question wording and timing to ensure that the interview was informal and not threatening to the respondents. To make sure ‘tha the flexible, semistructured interviews produced comparable or standardized in- {ormation trom each respondent, the imerviewers were given extensive training and field supervision. This approach, when combined with probability sampling of households in four carefully chosen communities, yielded precise quantitative i formation on Mexican migration to the United States ‘When the research purpose is not to derive facts or precise quantita scriptions, but to understand the meaning of respondents’ experiences (see 2002), unstructured interviews often are adopted. This approach allo Flexibility in the development of hypotheses and theory; however, it seldom is as= with survey research Hscause Tt invaRaDly involves small nonprobability samples and qualitative analysis An example is Kristin Luker's (1975) study ofthe relationship between contraception and abortion. By using a relatively unstructured approach in interviewing a sample of fifty abortion-secking women Who had expe- rience with effective methods of contraception but who had taken risks ding their last menstrual cycle, Luker was able to develop a theory of contaceptive tisk tak- ing. She found that « woman's decision not to contracept was nota symptom of psychological problems, as often has heen assumed by health eae professionals, but rather a rational decision based on the perceived costs and benefits tothe woman ‘of contraceptng and the perceived risks, cost, and benefits of pregnancy. Costs as- Suey Research as sociated with contraception included medical side effects, inconvenience, lack of spontaneity, male resistance, and so forth. To most of the women interviewed, the risk of pregnancy scemed slight, and all were aware ofthe possibilty of obtaining ‘an abortion, Benefits to getting pregnant included testing fertility (many were ofthe impression that they had some physical problem that decreased their chances of gt ‘ting pregnant) and esting the partner's commitment. For some research purposes, a social scientist might utilize two or three sets of interviews, beginning with very loosely structured interviews and progressing. to a final set of highly structured interviews. A. freee interviewing style in the pfeliminary stages would yield rich and varied information. This would assist the researcher in formulating or efining hypotheses, clarifying objectives, and spec~ ‘fying subtopies for subsequent semistructured interviews. Findings from a sec ‘ond et of interviews might be applied to the development ofa highly structured questionnaire Quantitative Data Analysis ‘Data analysis techniques depend on whether the survey's purpose is descriptive, ex planatory, ora combination of the two. Surveys that are primarily deseriptive seek to describe the distribution within « population of certain characteristics, attitudes, ‘or experiences and make use of simpler forms of analysis. Explanatory surveys, on the other hand, investigate relationships between two or more variables and altempt ‘explain these in cause-and-effect tems. Sorting out the relationships between the ‘variables in an explanatory survey requires the use of more sophistivated data anal {sis techniques. (Data analysis willbe dealt with further in chapters 14 and 15.) ‘An illustration of primarily descripave study isthe National Health and So- cial Life Survey (NHSLS) in which national probability sample of adults between the ages of 18 and 59 were interviewed about their sexual attiudes and prectces (Caumann etal, 1994). Originally prompred by the lack of national statistics on the prevalence of various sexual behaviors relevant tothe AIDS epidemic, NSLS was transformed by actions of federal agencies and hostile politicians ino a study with 1 somewhat broader focus, a much smaller than planned sample (3432 people were interviewed rather than 20,000), and a sponsorship by private rather than by public funding. : ‘Among the interesting NHSLS findings are estimates of same-gender sexual sctvity and forced sexual behavior. Only 1 percent of the women and 28 percent of the men during the interview identified themselves as “homosexual” or “bisex- tual.” Yet in a short selfadministered questionnaire that the respondent placed in a scaled envelope, 4.3 percent ofthe women and 9.1 percent of the men reported hav {ng engaged in a sexual act with someone of the same gender since puberty. These and other NHSLS estimates of homosexuality are substantially lower than reported in the Kinsey report and other studies that did not use probability sampling (Lau- ‘mann etal, 1994:286-90), Perceptions of unwanted sexual activity varied by gen der. Although only 2.8 percent of the men reported ever having forced a wornan to ddo something sexusl that she did not want to do, 21.6 percent of the women re- ported forced sex by a man, 24 eros of DATA COULTON “The aforementioned NELS represents explanatory survey research, The study ‘began with a baseline survey of studeas’ school and school-related experiences in the eighth grade and continued with follow-up surveys two, Four, six, and twelve years later. The general purpose of the NELS was to relate the eighth-grade expe- riences fo high school performance and, in turn, ro later achievements in life. The ‘wealth of data enables investigators to examine the impact of numerous factors— for example, students’ school experiences and activities, their relationships. with peers, parents, and teachers, and characteristics of their schools—on students” aca- {demic achievement, persistence in school, participation in postsecondary education, and career choices. Secondary Analysis of Surveys In mos of the examples presented thus far, each survey had a central topic, and the investigators were responsible for all phases ofthe research project, from the for- ‘ulation of ideas and research design to the collection, analysis, and presentation of the data, Although this was once the dominant approach, researchers today are ‘more apt to analyze survey data collected by some other person or agency than £0 ‘condict an ofginal survey themselves (Presser, 1984). The analysis of survey data by analysts other than the primary investigator who collected the data is call see~ ‘ondary analysis, Analyses of the NELS data by Beckett Broh and others represent this form of researc (One of the many advantages of this research strategy is that sample size can be increased greatly by combining data from several surveys. For example, in their study of the long-term effects of education, Herbert Hyman, Charles Wright, and John Shelton Reed (1975) dew together fifty-four surveys conducted between 1949 land 1971, involving 76,671 respondents, from three sources: the Gallup poll, the [National Opinion Research Center (NORC) of the University of Chicago, and the Survey Research Center at the University of Michigan. Not surprisingly, very strong positive relationship was found between amount of education and correct an swers fo knowledge questions, However, this was true not only for “academic”-type ‘questions taken from the arts and sciences, but for questions relating to current af fairs and popular culture as well. Furthermore, the relationship persisted over time, suggesting that one of the lasting effects of education is a lifelong openness t0 Jeaming of tendency to seek information ‘The eatliest application of secondary analysis was to census dat and then, be- inning inthe 1950s, to opinion-poll data. These data, however, were collected pri- marily for administrative and journalistic purposes. The major impetus for sec ‘ondary analysis came in the 1970s with the adveat of surveys designed expressly forthe purpose of making high-quality data availabe to the social science research community (Glenn, 1978), The first large-scale survey of this type, called the Gen- eral Social Survey (GSS), began in 1972 and was conducted annually (except for 1979, 1981, and 1992) until 1994, Each GSS involved personal interviews with bout 1500 respondents, drawn from a probability sample of the adult population ofthe United States. Starting in 1994 the GSS shifted to biennial surveys with tice ‘the normal sample size Survey Research ns In contrast 10 most surveys, which have a central topic, the GSS is eclectic, ‘with questions pertaining to a broad range of attitudes and behavior. A small por tion of each GSS includes a common core of questions that are replicated from year {o year to facilitate esearch on soeial trends. The remainder of each survey is de voted largely to topical and cross-national modules, which comprise blocks of ques Lions relating to special topics. The cross-national modules, which are sponsored by the International Social Survey Program (ISSP), ate included in national surveys in ‘over 30 collaborating countries, The Stier and Lewis-Epstein comparative study of work time preferences used the 1997 ISSP module on Work Orientation, ‘The 2002 GSS contained topical modules on computers and the Incernet, racial and ethnic prejudice, child mental-health sigma, patiipation in the ars, rust in doctors, quality of working life, non-wage compensation for workers, altruism, transition to adulthood, and sexual behavior, as well as two ISSP modiles, one on family, gender, and work and the other on social networks and support, for com- parative studies. The altruism modile included questions on empathy, blood dons ‘ion, charitable contributions, voluntarism, and spontaneous helping. The ISSP fam- lly, gender, and work module, which replicated some questions from 1988 and 1994, contained items on sex-role attitudes, work history before and after marriage and children, household division of labor, marital decision-making, and family and work stress. Data from the GSS are often used inthe teaching of sociology and other soci science courses, especialy statistics (see Babbie, Halley, and Zaino, 2003; Fox, 2003). Moreover, analyses of GSS data can be found in over 7000 books, aticles, ‘chapters, and dissertations (T. W. Smith, personal communication, 2003). We make extensive use of GSS data in chapters 14 and 15. In this chapter and the next, we will report some of the survey methods and materials used in producing these data so thatthe ryder miay become more familiar with ths important database ‘The Uses and Limitations of Surveys [Now that we have examined a wide range of surveys, we are in position to con- sider the uses as well asthe stengths and weaknesses of this approach. With the exception of the NELS, conducted by a federal government agency, all ofthe stud ies cited above were carried out by socal scientists for social scientific purposes. Infact, “surveys are the most widely used method of collecting data in the social sciences, especially in sociology and political science” (Bradburn and Sudmai 1988:61). Yet the use of surveys asa scientific tool ouside the scientific comma. nity is even more extensive. The ubiquitous opinion pols, oft reported in the news media and eyed warily by politicians, monitor public reactions to people, evens, and policies. Marketing research by businesses, avertsing agencies, and other of. {ganizations tests consumer reactions to new produets and services, assesses cus- tomer satisfaction, and compiles audience profiles for various media, And the sin aie largest user of surveys, the federal government, conducts or commissions scores of surveys every year o help in planning, decision making, and policy assessment, Given the importance of surveys in shaping major decisions by politicians, busi- * 265 ‘MEMOS OF Dara coupe nesspeople, and government officials, we all need to know something about surveys (Bradburn and Sudman, 1988:2). What ean they tellus beter than other methods ‘of social research? And what are their major imitations? ‘Whereas experiments are used almost exclusively for explanatory, hypothesis- testing research, surveys are used extensively for both descriptive and explanatory purposes. Among all approaches to social research, infact, surveys offer the most effective means of social description; they ean provide extraordinarily detailed and precise information about large heterogeneous populations. By using probability Sampling, one can be certain, within known limits of error, whether the responses fo-a sample survey accurately describe the larger target population. Furthermore, the topics covered and the questions that may be included in surveys are wide rang ing. Topics ofthe studies cited above ranged from academic achievement t0 alco hol consumption and from sexual activity to attitudes toward capital punishment ‘The scope of possible survey question is suggested by the following classifica (Schuman and Kalton, 1985) 1, Social background information (eg, What is your tligious preference? What is your date of bit?) 2, Reports of past behavior e., Did ou vow in the ast presidential election? Fave you ever been te victim of a crime? On an average day, about how many hours do you personally watch television?) 3. Atitues. elif, and values (eg, Do you bolieve that there isa ife after ‘death Do you think there should be laws against mages between backs and whites?) 4. Behavior intentions (eg. I he presidental election were Bold today, whom ‘would you vote for? Would you Youre have an aboron if there is & song chance of serous defect in the baby?) 5. Sensitive questions (eg, Have you ever been arested fora crime? Have you sed cocaine in the past mont?) For categories 1, 2, and 5, which pertain to behavior and personal characteristics, the information may be verifiable from records or observer repors, but itis often impractical, unethical, o even illegal to obtain it from sources other than the indi- viduals themselves (Bradburn and Sudman, 1988). For subjective phenomena such as categories 3 and 4 the information can be directly known, if at all, only by ask ing the individuals themselves (Turner and Martin, 1984), 'As this listing suggests, surveys can address a much broader range of research topics than experiments can, Ethical considerations preclude studying some topics ex- petimentally—for example, the effect of emotional traumas on mental health —while practical considerations rule out many others; fr instance, one nonmally cannot ex- perimentally manipulate organizations ot nations. Besides this flexibility, surveys can be every efficient dara-gatering technique. While an experiment usually wil adiess ‘only one research hypothesis, numerous research questions can be jammed into a s fle large-scale survey. Furthermore, the wealth of data typically contained ina com pleted survey may’ yield unanticipated findings or lead to new hypotheses. "The secondary analysis of surveys also affords many unigue advantages. The cost of obtaining the data for analysis is usually @ small fraction of the cost of co Survey Research 27 Jectng and coding the data, Survey data made avaiable for secondary analysis tend to come from professional polling and research cenlers with the resoures to obtain igh-qualy information from large, national samples, In ation, secondary sna yrs may enable one to (a) astess social trends by examining questions repeated Overtime and () increase sample size by combining data fom several surveys. ‘The major disadvanage of surveys eats to dei use in explanatory research Beyond association between variables, the citeia for inferring cause-andeectre- lationships cannot be established as easily in surveys as in experiments. For exam- ple, the eietion of diretonaiy—that & cause mus inuence is effect—ispre-

    You might also like