Jump to ContentJump to Main Navigation
Soul SearchingThe Religious and Spiritual Lives of American Teenagers$

Christian Smith and Melinda Lundquist Denton

Print publication date: 2005

Print ISBN-13: 9780195180954

Published to Oxford Scholarship Online: October 2005

DOI: 10.1093/019518095X.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017

(p.292) Appendix B. Survey Methodology

(p.292) Appendix B. Survey Methodology

Source:
Soul Searching
Publisher:
Oxford University Press

The National Survey of Youth and Religion (NSYR) is a nationally representative telephone survey of 3,290 English- and Spanish-speaking teenagers between 13 and 17, and of their parents. The NSYR also includes 80 oversampled Jewish households, not nationally representative (described below), bringing the total number of completed NSYR cases to 3,370. The survey was conducted from July 2002 to April 2003 by researchers at the University of North Carolina at Chapel Hill using a random-digit-dial (RDD) method, employing a sample of randomly generated telephone numbers representative of all household telephones in the 50 United States, including Alaska and Hawaii. The national survey sample was arranged in replicates based on the proportion of working household telephone exchanges nationwide. This RDD method ensures equal representation of listed, unlisted, and not-yet-listed household telephone numbers. Eligible households included at least one teenager between 13 and 17 living in the household for at least six months of the year.1 To randomize responses within households, and so help attain representativeness of age and gender, interviewers asked to conduct the survey with the teenager in the household who had the most recent birthday. Parent interviews were conducted with either a mother or father, as available, although the survey asked to speak with mothers first, believing that they may be better qualified to answer questions about their family and teenager. Stepparents, resident grandparents, resident partners of parents, and other resident parent-like figures were also eligible to complete the parent portion of the survey.

An RDD telephone survey sampling method was chosen for this study because of the advantages it offers compared to alternative survey sampling methods. Unlike school-based sampling, for example, our RDD telephone method was able (p.293) to survey not only school-attending youth, but also school dropouts, home-schooled youth, and students frequently absent from school. Using RDD, we were also able to ask numerous religion questions which many school principals and school boards often disallow on surveys administered in school. Explicit informed consent from parents also proved more feasible using RDD than school-based sampling. And the verbal reading of survey questions by trained interviewers facilitated question-and-answer clarifications that increased the validity of answers, compared to paper-and-pencil questionnaires administered en masse in school classrooms. Also, given the relatively low incidence rate (14 percent) of American households with teenagers 13 to 17 years old, the NSYR's RDD telephone survey method was much more cost-effective than an in-home survey, which would have been cost-prohibitive. The NSYR's RDD telephone method also eliminated potential design effect problems associated with sampling from a limited number of geographic or school clusters. Furthermore, the greater anonymity of an RDD survey interviewer on the telephone, compared to an in-person interviewer in the home, may have also increased the validity of teenagers' answers to sensitive questions and reduced possible biasing effects of in-person interviewers' sex, race, and age.2 No good sampling frames exist with which the NSYR might have conducted a mail survey, which typically garners extremely low cooperation and response rates in any case. Finally, superior Internet-based methods of sampling and surveying were not sufficiently developed and tested by the time of this survey's fielding to have been useful for the NSYR.3

Prior to conducting this survey, the researchers conducted 35 in-depth pilot interviews with teenagers to help inform the construction of the survey instrument. The researchers also conducted additional survey-focused interviews and focus groups with a variety of types of teenagers to improve question wording and comprehension. Prior to the survey, researchers also conducted modest (N=175) pretests of the survey instrument using both nationally representative and convenience samples. Based on pretest results, the researchers revised questions and answer categories to enhance survey clarity and validity. The final survey instrument is available by Internet download at the project Web site: www.youthandreligion.org/publications/docs/survey.pdf.

The NSYR survey was conducted with members of both English- and Spanish-speaking households. The English version of the survey was translated into Spanish by a professional translation service. That translation was then closely reviewed, evaluated, and revised by four separate native Spanish-speaking translation consultants and six Spanish-speaking survey interviewers to ensure the best translation for Spanish-speaking respondents. The final Spanish-language version was then programmed into the computer-assisted telephone interview (CATI) system for calls to Spanish-speaking households. Surveys with Spanish-speaking households were conducted by native Spanish-speaking interviewers who are fluent in both English and Spanish and who had extensive experience conducting the survey in English before conducting the Spanish version. The parent and teen respondents from households could each choose the language with which to complete the survey, so that a parent might use the Spanish version, for example, while the teen used the English version. Spanish-speaking household numbers are included in the calculations of the national sample cooperation, completion, and response rates below.4

All survey interviewers received two days of project-specific training in the (p.294) significance and purpose of the survey, the meaning of all survey questions and their answer categories, the proper pronunciation of religious terms, and the ethical treatment of human subjects. They also completed an Internet-based Human Participant Protections Education for Research Teams course offered by the National Institutes of Health, U.S. Department of Health and Human Services (www.cme.nci.nih.gov). Immediately prior to conducting all surveys, interviewers obtained respondents' verbal informed consent and provided respondents with information about the confidentiality of their answers and their right to refuse to answer questions. Household eligibility was determined through the use of an initial screening question about resident teenagers. Incentives of $20 to parent respondents and $20 to teenage respondents were offered to complete the survey, for a total of $40 to completing households.5 Survey respondents were also able to complete the survey at their convenience by calling a toll-free number that linked to their sample record. Throughout the fielding of the survey, interviewers were monitored by project staff (primarily the authors) using remote technology to ensure data quality, and the interviewers, monitors, and researchers were routinely debriefed about survey performance. Upon completing the survey, all respondents were given contact information for the researchers, the research firm, and the university Institutional Review Board to verify the survey's authenticity or ask any questions about the survey or their rights as respondents. This information was also included in written form in thank-you letters accompanying the mailed incentives.

To help protect the privacy of survey respondents, the NSYR obtained a Federal Certificate of Confidentiality from the National Institutes of Health. With this certificate, researchers with the NSYR could not be forced to disclose information that might identify respondents, even by a court subpoena, in any federal, state, or local civil, criminal, administrative, legislative, or other proceedings. The certificate was thus useful for resisting any potential demands for information that would identify respondents (with the following exceptions: a Certificate of Confidentiality does not prevent respondents or members of their families from voluntarily releasing information about themselves or their involvement in the NSYR; if and when an insurer, employer, or other person were to obtain respondent's own voluntary written consent to receive research information, the NSYR could not use the certificate to withhold that information; neither does the Certificate of Confidentiality prevent the researchers from disclosing without respondents' consent information that would identify them as a participant in the research project in stated cases of child abuse or intent to hurt self or others; if teen respondents disclosed evidence of neglect or abuse, the NSYR had an obligation to inform the appropriate authorities).

The NSYR survey was conducted over nine months, between the end of July 2002 and the beginning of April 2003. All randomly generated telephone numbers were dialed a minimum of 20 times over a minimum of five months per number, spread out over varying hours during weekdays, weeknights, and weekends. The calling design included at least two telephone-based attempts to convert refusals. Households refusing to cooperate with the survey yet established by initial screening to have children age 13 to 17 in residence and with telephone numbers able to be matched to mailing addresses were also sent information by mail about the survey, contact information for researchers, and a request from the principal investigator to cooperate and complete the survey; those records were then called (p.295) back for possible refusal conversions.6 Most cell phone numbers were screened out of the initial sample through the identification of unique cell phone exchange numbers. All other nonhousehold numbers (business, government, nonprofit, payphones, remaining cell phones, etc.) were screened out of the sample through direct calling dispositions and ascription of contact and noncontact telephone numbers for noncompletes based on proportions of household numbers among working telephone numbers.7 Of the sample producing only voice-mail or answering machine, 34 percent were dialed more than 99 separate times; 23 percent of those voice-mail numbers were dialed between 50 and 99 times, and 43 percent of those voice-mail numbers were dialed between 25 and 49 times.

The NSYR survey itself took a mean of 30 minutes to complete the parent portion and 52 minutes to complete the teen portion of the instrument, for a mean parent-teen combined survey length of 82 minutes. A total of 3,370 respondent households completed our full survey, 3,290 of which were RDD national sample respondents and 80 of which were Jewish oversample respondents. The overall cooperation rate of our national sample was 81 percent. Ninety-six percent of parent complete households also achieved teen completes. Using the American Association for Public Opinion Research (AAPOR) RR4 calculator, the final NSYR national sample survey response rate was 57 percent.8 For descriptive purposes, a weight was created (“weight2”) to adjust for number of teenagers in household, number of household phone numbers, census region of residence, and household income.9

Diagnostic analyses demonstrate that the NSYR provides a nationally representative and unbiased sample of U.S. households with resident teenagers age 13–17. Comparative tests were run for the national representation of or potential sampling biases in the NSYR sample employing known population characteristics on key variables from 2002 U.S. Census data. Table B.1 shows that the NSYR provides a nearly perfectly representative sample of 13- to 17-year-olds living in U.S. households by the comparable variables of gender, age, race/ethnicity, and household type. The region and household income variables also demonstrate a very close representation by NSYR data of the known national population, which is nearly entirely corrected for when weights are applied.10

Comparisons are also made in table B.2 between weighted NSYR data and 1999 National Household Education Survey (NHES) data, 1996 Monitoring the Future survey data, 1994 National Longitudinal Survey of Adolescent Health (Add Health) data, and 1999 Survey of Adults and Youth data on key comparable variables. Table B.2 reveals only minor differences between the NSYR and other samples on these variables. The percentage in the NSYR sample who never drink alcohol, never used marijuana, and who smoke cigarettes regularly is similar to the percentages found in the Monitoring the Future and Add Health surveys. Based on these interdataset analyses, we can say with some confidence that findings from the NSYR appear to offer a reasonably unbiased representation of the sampled population and so, particularly when region and income are weighted, might be assumed to accurately describe the population of U.S. teenagers age 13–17 and their parents living in residential households.

These between-dataset comparisons were supplemented with analyses in the NSYR dataset comparing key demographic and behavioral traits of respondents who initially cooperated (90 percent of the sample) with respondents who were initial refusals but subsequently cooperated after successful attempts at refusal (p.296)

Table B.1. Demographic Characteristics Comparing Unweighted and Weighted NSYR Samples and the 2002 U.S. Census Population, Households with 13- to 17-Year-Olds in Residence, 2002–3 (Percentages)

NSYR

2002

NSYR

(Unweighted)

Census

(Weighted)

Census Region

Northeast

15

17

17

Midwest

23

22

22

South

42

37

37

West

20

24

24

Gender

Male

50

51

50

Female

50

49

50

Age

13

19

20

19

14

19

20

20

15

21

20

21

16

20

20

21

17

20

20

20

Teen race/ethnicity

White/Hispanic

77

78

78

Black

17

16

16

Asian/Pacific Islander/American Indian/mixed/other

5

7

5

Household type

Married couple

67

68

70

Income

Less than $10K

4

6

5

$10K–$20K

7

9

10

$20K–$30K

13

10

10

$30K–$40K

14

11

11

$40K–$50K

14

11

11

$50K–$60K

12

8

8

$60K–$70K

8

8

9

$70K–$80K

7

7

8

$80K–$90K

5

6

6

$90K–$100K

4

5

5

More than $100K

13

19

19

conversion (10 percent of the final sample).11 Results, displayed in table B.3, reveal that the refusal-conversion respondents in the sample disproportionately are somewhat more Midwestern (and less Southern), involve more male parent and teen respondents, involve more 15- and 16-year-old teen respondents (and fewer 13- and 14-year-old respondents), are white (and not black), represent more married parent households, represent higher-income families, and represent more regular religious service–attending parents. Most of these differences are quite modest, however, typically representing less than 10 percentage point spreads. Nevertheless, most of these differences remain statistically significant when considered together in logistic regression analyses predicting refusal-conversion versus (p.297)

Table B.2. Comparison of Weighted NSYR Results with Parallel National Survey Results on Selected Lifestyle Variables, U.S. Adolescents, Ages 13–17 (Percentages, unless otherwise noted)

NSYR

Monitoring the Future

Add Health

Survey of Adults and Youth

NHES

(N=3,290)

(N=45,173)

(N=15,084)

(N=874)

(N=6,569)

(2002–3)

(1996)

(1996)

(1999)

(1996)

Never drinks alcohol

63

62

61

Never uses marijuana

75

64

74

Smokes cigarettes regularly

7

9a

9

Nights per week eats dinner with one parent (mean)

5

5

5

School type attending

Public

87

90

Private religious

7

7

Private secular

2

2

Home-schooled

2

2

Attends religious services weekly or more

41

33

39

45

47

Never attends religious services

18

15

14

11

13

initial-cooperator survey respondents (results not shown). These modest differences suggest that, when conceptualizing a survey respondent continuum from easy cooperators to serious noncooperators—and differences between potential respondents who lie on that continuum—successful researcher efforts to convert initial refusals for inclusion in the final dataset increase the full representation of different types of respondents in the final sample, reducing possible sampling biases that would be found in surveys with less rigorous refusal-conversion methods. Because the NSYR successfully employed multiple, extensive, sustained measures to convert initial refusals into cooperators—who represent fully 10 percent of the final sample—it significantly reduced possible biases affecting measured variables potentially associated with respondents' propensity to cooperate with surveys, rendering its sample data more accurately representative. This assurance, together with the between-dataset analyses presented above, corroborates this methodological report's general conclusion that the NSYR provides a satisfactorily unbiased, nationally representative sample of the target population of U.S. teenagers age 13–17 living in households.

In addition to a main national sample of 3,290 cases, the NSYR conducted surveys with a modest oversample of Jewish households, 80 completes in all, to help obtain a large enough number of cases with which to conduct meaningful statistical analyses of Jewish youth. A national RDD sampling method screening out all but eligible Jewish households with teenagers would, theoretically, have been ideal, insofar as it would have provided a true probability sample yielding a genuinely nationally representative oversample of Jewish teenagers in propor (p.298) (p.299)

Table B.3. Comparison of Weighted NSYR Initially Cooperating and Refusal-conversion Respondents (Percentages)

Initial Cooperators

Refusal Conversions

(N=3,030)

(N=340)

Census region

Northeast

17

17

Midwest

21

28

South

38

32

West

24

23

Male parent

18

26

Male teen

50

54

Teen age

13

19

15

14

20

16

15

21

23

16

20

26

17

20

20

Race/ethnicity

White/Hispanic

78

81

Black

16

13

Asian/Pacific Islander/American Indian/mixed/other

6

7

Household type

Married Couple

69

77

Income

Less than $10K

5

3

$10K–$20K

10

4

$20K–$30K

9

8

$30K–$40K

11

9

$40K–$50K

10

9

$50K–$60K

8

7

$60K–$70K

9

9

$70K–$80K

7

6

$80K–$90K

5

7

$90K–$100K

4

7

More than $100K

16

27

Parental religious service attendance

Parent attends weekly or more

44

50

Parent never attends

15

13

tion to their actual geographic and social locations in the United States. However, because Jewish households with resident teenagers age 13–17 represent only about 1 in every 400 U.S. households (approximately 0.25 percent of all households), this method would have been unreasonably time-consuming and cost-prohibitive. As a more efficient alternative, the NSYR employed another standard survey method for obtaining a Jewish oversample, by calling a set of telephone numbers listed with one of 200 “Jewish” surnames agreed on by the National Jewish Technical Advisory Committee and selected from White pages listings throughout the United States on a population-proportional basis. These numbers were obtained from the survey sampling firm Genysis, Inc., and screened in calling for eligible households with resident 13- to 17-year-olds. Compared to yet a third alternative oversampling method—namely, the high-density Jewish sampling frame method of RDD calling of replicates of telephone numbers within geographic areas containing defined minimum Jewish residency rates, which, by definition, samples areas with higher concentrations of Jewish inhabitants, schools, synagogues, and other Jewish institutions and produces all of the associated sampling biases—the listed Jewish surname method employed by the NSYR has the distinct advantage of sampling Jewish youth from all social and geographic locations in the country. Nevertheless, the listed surnames sample method employed is by no means a nationally representative probability sample, because it systematically excludes both Jewish households with unlisted telephone numbers and Jewish households with surnames not considered by the survey sampling firm to be “Jewish-sounding.”

To estimate any possible sampling bias involved in this nonprobability oversampling method, analyses compared key characteristics of the Jewish cases drawn from the nationally representative sample to those drawn from the Jewish oversample on a variety of key demographic and religious measures. Cross-tabular comparisons of the unweighted Jewish national sample and oversample cases reveal no statistically significant differences with regard to the following variables: teen gender, age, race, religious service attendance, religious youth group participation, belief in God, importance of religious faith, type of school attending, household income, family debt or savings, family home ownership, parental marital status, parental religious service attendance, parental importance of faith, and father's education.12 The two differently sampled groups, then, are remarkably similar along important dimensions of analysis. Therefore, it is deemed not necessary to construct or use special weights to compensate for any religious or demographic differences generated by the nonprobability oversampling method used here.

In sum, the National Survey of Youth and Religion may be taken as providing a nationally representative RDD telephone survey of 3,290 English- and Spanish-speaking teenagers living in households in all 50 U.S. states in the years 2002 and 2003, between the ages of 13 and 17, and of their parents; the survey also includes 80 oversampled Jewish households (not nationally representative), bringing the total number of completed NSYR cases to 3,370. Every effort was made in project design, instrument construction, interviewer training, and survey fielding to produce the best possible results. Multiple diagnostic analyses demonstrate that the NSYR appears to provide a reasonably unbiased representative sample of its target population and so, when weights are applied, can be taken to accurately describe the population of U.S. teenagers age 13–17 and their parents living in residential households. (p.300)

(p.301)

Table B.4. Demographic and Religious Traits of Religious Ideal Types of U.S. Adolescents, Ages 13–17 (Percentages)

U.S.

Religious Type

Devoted

Regulars

Sporadics

Disengaged

Age

13

19

16

20

18

16

14

20

23

21

22

19

15

21

21

21

19

18

16

21

24

20

18

24

17

20

17

18

23

24

Female

49

63

46

45

40

Race

White

66

78

67

69

80

Black

17

10

13

10

5

Hispanic

12

9

12

14

7

Other race

5

3

8

7

8

Region

Northeast

17

10

17

22

21

Midwest

22

19

28

24

23

South

37

46

33

28

21

Mountain/Pacific

24

25

23

26

35

Parent marital status

Married

70

81

74

70

67

Living with unmarried partner

4

4

3

6

7

Widowed

2

2

2

2

3

Divorced

13

9

12

12

16

Separated

5

2

5

3

3

Never married

6

3

5

7

4

School type attending

Public

87

80

89

90

88

Catholic

4

1

6

4

2

Private other Christian

3

12

3

1

Jewish

Private nonreligious

2

2

1

2

3

Home-schooled

2

4

1

1

4

Other or not going to school

2

1

3

2

Religious type

Conservative Protestant

30

59

30

23

6

Mainline Protestant

11

14

14

13

5

Black Protestant

11

8

10

8

1

Catholic

27

4

37

39

16

Jewish

2

1

5

3

Mormon

3

12

3

2

2

Other religion

5

4

5

7

3

Not religious

12

3

65

Table B.4 shows the demographic characteristics of the four religious ideal-type categories used in the analysis of chapter 7.

Notes:

(1.) Another survey conducted using similar methods is the 1998–99 Survey of Parents and Youth (SPY, later renamed the Survey of Adults and Youth [SAY]), which was designed by Princeton University's Center for Research on Child Wellbeing in conjunction with the National Evaluation Team for the Urban Health Initiative at the (p.338) Center for Health and Public Service at New York University's Robert F. Wagner Graduate School, and was funded by the Robert Wood Johnson Foundation. SPY was designed to monitor trends in youth's access to parental and community resources and included interviews with parents and youth. SPY was administered as an RDD telephone survey to a nationally representative sample of youth age 10–18 and to oversamples of youth in five selected cities (Philadelphia, Baltimore, Detroit, Oakland, Richmond, and Chicago). Parents were screened and then interviewed, after which point the interviewers asked permission to interview the youth. SPY was conducted in English, Spanish, or Chinese, and lasted an average of 30 minutes for youth and 20 minutes for parents.

(2.) For some teenagers, to be sure, the anonymity of a telephone survey may have increased their level of discomfort with sensitive questions; we cannot know with certainty.

(3.) See Melinda Lundquist Denton and Christian Smith, Methodological Issues and Challenges in the Study of American Youth and Religion, project report (Chapel Hill, NC: National Study of Youth and Religion, 2001). One disadvantage of an RDD-sampled survey is that it does not include the approximately 4 percent of U.S. households without telephones at any given time. This concern, however, is somewhat alleviated by the fact that the majority of households without telephone service are not permanently so, but typically fluctuate in and out of having service over time, thus increasing their chances of inclusion in this survey insofar as it was conducted over seven months, and by the likelihood that households with teenagers in residence are underrepresented among households lacking telephone service. Neither does the NSYR represent those households with teenagers 13–17 whose telephone service consists only of cell phones, of which we would expect there to be very few.

(4.) See Leo Morales, “Cross-Cultural Adaptation of Survey Instruments,” in Assessing Patient Experiences with Assessing Healthcare in Multi–Cultural Settings (Santa Monica, CA: Rand Corporation, 2001); R. W. Brislin, “The Wording and Translation of Research Instruments,” in W. J. Lonner and J. W. Berry, eds., Field Methods in Cross-Cultural Research (Beverly Hills, CA: Sage, 1986).

(5.) Such incentives increase response rates without appearing to distort the quality of responses. See Eleanor Singer, “The Use of Incentives to Reduce Nonresponse in Household Surveys,” in Robert Groves, Don Dillman, John Eltinge, and Roderick Little, eds., Survey Nonresponse (New York: Wiley, 2002). In the case of our (mean) 82-minute-long survey, such incentives seemed particularly important.

(6.) Such extensive and persistent contact efforts are expected significantly to reduce nonresponse bias. See Peter Lynn, Paul Clarke, Jean Martin, and Patrick Sturgis, “The Effects of Extended Interviewer Efforts on Nonresponse Bias,” in Robert Groves, Don Dillman, John Eltinge, and Roderick Little, eds., Survey Nonresponse (New York: Wiley, 2002).

(7.) See Federal Communications Commission, Statistics of Communications Common Carriers (Washington, DC: Federal Communications Commission, 2000); J. Michael Brick, Jill Montaquila, and Fritz Scheuren, “Estimating Residency Rates for Undetermined Telephone Numbers,” Public Opinion Quarterly 66 (2002): 18–39. Survey noncooperators are also known to have a relatively higher incidence of single-person and elderly households and of households without children in residence. See Robert Groves and Mick Couper, Nonresponse in Household Interview Surveys (New York: Wiley, 1998), pp. 119–154.

(8.) AAPOR RR4 is calculated thus: (I+P)/(I+P)+(R+NC+O)+e(UH+UO), where (p.339) I = completed interview, P = partial interview, R = refusal and breakoffs, NC = noncontacts, O = other contacts, e = estimated proportion of cases of unknown eligibility that are eligible, UH = unknown if household occupied, and UO = other unknown; see www.aapor.org under “Survey Methods: Response Rate Calculator.” A 90 percent household rate is ascribed for voice mail, unknown qualification callback, and unknown qualification refusal sample; among households, a 4 percent teen qualified incidence rate is ascribed for voice mail, unknown qualification callback, and unknown qualification refusal sample. Calculated with no adjustments for teen qualification incidence rates (e), thus more conservatively, the response rate is 48 percent. Comparative studies show an annual negative trend in survey response rates across Western nations due to increased difficulties with respondent contact and cooperation. See Edith de Leeuw and Wim de Heer, “Trends in Household Survey Nonresponse: A Longitudinal and International Comparison,” in Robert Groves, Don Dillman, John Eltinge, and Roderick Little, eds., Survey Nonresponse (New York: Wiley, 2002). While a higher response rate is always desirable, the NSYR's final response rate represents among the highest possible given the methodology employed and prevailing social conditions and available technologies. Findings of extensive comparative analyses of the data (described in the next paragraph) document the NSYR's lack of sampling bias and national representation of U.S. teenagers 13–17, significantly mitigating response rate concerns.

(9.) See Jelke Bethlehem, “Weighting Nonresponse Adjustments Based on Auxiliary Information,” in Robert Groves, Don Dillman, John Eltinge, and Roderick Little, eds., Survey Nonresponse (New York: Wiley, 2002); Judith Lessler and William Kalsbeek, Nonsampling Error in Surveys (New York: Wiley, 1992), pp. 183–193; C. H. Fuller, “Weighting to Adjust for Survey Nonresponse,” Public Opinion Quarterly 38 (1974): 239–246. The 6 percent of cases with missing data on income, a much lower number than is typical of survey income data, were imputed using these variables in the following order of assigned importance: resident father's and mother's education, parental marital status, family home ownership status, and family race.

(10.) The South was overrepresented and the Northeast underrepresented by 5 and 2 percentage points, respectively; incomes greater than $100,000 were underrepresented by 6 percent.

(11.) Here we follow a similar procedure employed with success by the Survey of American Attitudes and Friendships; see Michael O. Emerson, “Report on the Lilly Survey of American Attitudes and Friendships,” Department of Sociology, Rice University, Houston, 2000; also see Michael O. Emerson, George Yancey, and Karen J. Chai, “Does Race Matter in Residential Segregation? Exploring the Preferences of White Americans,” American Sociological Review 66, no. 6 (2001): 922–935.

(12.) On only two measures tested were the two groups different: (1) the parent respondent for the oversample cases was 6 percent more likely to be a mother than a father; and (2) the educational attainment of the mothers of the oversample respondents was greater (only 5 percent of them had not attended any college, compared to 14 percent of the national Jewish sample mothers).