Introduction
The well-known changes in American families over the past few decades have greatly increased the percentage of children who do not reside with their fathers. In 1970, 85 percent of all children under 18 were living with both their parents, whereas by 1995, only 69 percent were doing so. Another 23 percent of children lived with their mother only, 4 percent with their father only, and 4 percent lived with neither parent (U.S. Bureau of the Census, Department of Commerce).
As the composition of families has changed, much attention has focused on the roles of absent fathers in their children's lives. At first the focus was on the economic contributions of these fathers. In recent years, public concern has been wide-ranging, encompassing the psychological, social, educational, and health consequences of absent fathers. Moreover, as men's family roles have changed, the family and fertility behavior of all fathers, present as well as absent, has become of greater interest to researchers. Yet social scientific evidence on the process of becoming a father and on what fathers do is limited.
In addition, as parenthood has become decoupled from marriage, the reproductive careers of men have become more distinct from the reproductive careers of women. Men's sexual activities encompass a greater number of partners over the life course than was the case a few decades ago. Because of the increases in divorce and childbearing outside of marriage, men are more likely to have had children by two or more women than was the case a generation or two ago. The rise of cohabitation has led to informal partnerships that are sometimes of short duration.
To be sure, these trends have affected women similarly. But for a number of reasons, men's reproductive careers have the potential to be more complex than women's. Men are not limited by pregnancy and they typically do not provide primary care for young children; moreover, their rates of remarriage after divorce are higher than women's (Cherlin, 1992). Consequently, they report more sexual partners than do women, (Laumann, Gagnon, Michael, and Michaels, 1994) and they are more likely to produce children with multiple partners than are women. Since they are not likely to be living with children from previous unions, they may underreport the existence of those children. It is, therefore, a greater challenge to obtain complete information about sexual, reproductive, are union histories of men than of women.
What is more, we know much less about becoming and being a father than we do about becoming and being a mother. Since 1955, American demographers have fielded a series of surveys of the fertility of women. In 1973, the Federal government took over responsibility for the series, which it entitled the National Survey of Family Growth. It comprises a complex and sophisticated survey of women of childbearing age. Recent waves have asked about the men in the women's lives. But it does not include interviews with men.
In fact, the overwhelming majority of social scientific studies of children's family lives have focused on mothers rather than fathers, even when the fathers were present in the home. Perhaps the fundamental methodological problem that we face in studying fathers is that the household survey, the basic data gathering tool for demographic and behavioral science research on the family, the labor force, and fertility, was designed based on assumptions that no longer hold. When the standard household survey was being developed at mid-century, it was reasonable to assume that a family lived in just one household. The divorce rate and the percentage of births outside of marriage were far lower than they are as we approach the twenty-first century. Thus, it was reasonable to assume that complete and accurate information about a family unit could be obtained from a single household.
However, social change has undermined this assumption. Increasingly, families extend across the boundaries of households, so that the standard survey, focused on the members of one household, is no longer a sufficient method for obtaining complete and accurate information about family relationships. It is obvious that the standard household survey is deficient in providing complete and accurate information about non-resident fathers. It is less obvious but still true that the standard survey--focused as it is on mothers and children in the household--is deficient in providing a complete sexual, reproductive, and union history of men in the household.
As a result, best-practice studies of fathers and families have already moved beyond the standard survey practices of mid-century. Currently, a number of methodological innovations in survey research are being developed and tested. We will describe many of these below. This line of methodological research is still new, and much more work is needed. We applaud this line of research and call for its expansion.
Survey-based studies, however, are inherently limited in the kind of information they can provide. Surveys are best used as hypothesis-testing mechanisms after a general understanding of a topic has been obtained. But when little is known about the behavior of interest, as is the case with father-child relations, surveys cannot provide a full picture. Rather, more intensive studies are necessary as hypothesis-generating mechanisms. These studies include the intensive observation that developmental psychologists specialize in and ethnographic studies of the kind practiced in anthropology and sociology. We endorse further use of these methods also.
Other Working Groups at the Fatherhood Conference being sponsored by the National Institute for Child Health and Human Development will address the substantive issues concerning fathers in considerable detail. Underlying these substantive questions are important methodological issues that must be addressed before we can have confidence in data to be collected on fathering and fatherhood. The Working Group on the Methodology of Studying Fathers was established to address these issues, in consultation with other working groups.
The organization of this paper is as follows. Section 1 reviews current studies that provide some information about fathers or that have interesting methodological approaches that yield insights into improving data on fathers. Section 2 comprises a lengthy examination of a number of methodological issues that are related to the quality and characteristics of data on fathers. Section 3 examines the issue of how new data collection should be undertaken. Section 4 presents our recommendations.
Current Activities
Before discussing current methodological issues, let us briefly summarize some of the major national surveys with protocols that are of methodological interest. While the debate evolves in the statistical and research communities as to what information is needed and how it needs to be collected, important initiatives are being made in both publicly and privately sponsored surveys. A brief overview of major activities and studies that are currently the primary sources of information on fathers serves to inform the discussion about further advances that may be required.
Studies of Methodological Interest (listed alphabetically):
Add Health. Add Health, a national longitudinal study of adolescent health, is a comprehensive study of the health and health behaviors of adolescents that has been uniquely designed to measure the contextual factors that influence these outcomes. Outcomes to be examined include behaviors related to fertility as well as a broad range of other health-related behaviors and outcomes; antecedents include measures of adolescents' relationships with their resident and nonresident fathers. The study features a longitudinal, multi-level design with independent measurement at the individual, family, peer group, school, and community levels; further, the study is designed to provide information from both partners to romantic relationships in a substantial number of cases. The basic sample is drawn from a stratified probability sample of 80 high schools and 80 feeder schools (middle or junior high schools) nationwide. Information on peer networks, nonsensitive health behaviors, and school climate is collected in the schools from all students attending grades 7-12. Subsequent interviews are conducted in individuals' homes with a subsample of 20,000 adolescents drawn from the school rosters and with a parent of each adolescent. Adolescents are re-interviewed after one year. All adolescent interviews are conducted with a laptop computer, with sensitive portions of the interview self-administered via audio-CASI.
Early Childhood Longitudinal Study (ECLS). The ECLS will be a national, longitudinal cohort of kindergartners in fall 1998, to be followed once or twice a year through at least fifth grade. The study is sponsored by the National Center for Education Statistics. The household roster will obtain some information about persons who have lived with the child at some point in the past for four months or more. Detailed interviews will be conducted with the child, the mother, teachers, and school administrators over the life of the study. There are no current plans to interview the fathers or to seek to find absent parents and interview them.
National Adult Literacy Study (NALS). NALS, sponsored by the National Center for Education Statistics and conducted in 1992, is one of the few national sample surveys that combined a household sample of the noninstitutionalized population with a national sample of inmates in state or federal prisons. Nearly 1,150 inmates in 80 federal and state prisons were interviewed and tested for their literacy skills. These respondents were included in both a separate data set and in national population estimates. This proved an effective strategy for providing a more comprehensive look at the literacy skills of a larger segment of the population. The inclusion of inmates may help to address undercoverage in surveys of fathers.
National Household Education Survey (NHES). The NHES is a random-digit-dial telephone survey that uses computer-assisted telephone interviewing technology to collect data on high priority topics that cannot be addressed adequately through school- or institutional-based surveys. The 1996 NHES included a parent involvement component that asked the parents/guardians of 16,910 kindergartners through 12th graders questions about mothers' and fathers' involvement in their children's schools. The survey also asked about children's contact with nonresident fathers and about the involvement of these fathers in their children's schooling. Responses were provided by the resident parent, usually the mother. The sample included 5,440 children who had a nonresident father.
National Longitudinal Survey of Youth, 1997 (NLSY97). The NLSY97 will attempt to roster all people living in the residence of the sample youth as well as relatives who live outside of the household including biological, adoptive, and step-parents; full, half, and step siblings; non-resident children of parents in the household, and the other parent of any such children. Information solicited about these people will depend on the relationship of the sample youth to the person. While address information will be obtained when possible for absent parents, there are currently no firm plans for follow-up with absent parents. The survey will include a parent interview in the initial year, and could have additional parent surveys in later years.
National Survey of Adolescent Males. Since 1988, three waves of this study have interviewed young men about their sexual, contraceptive, and HIV-prevention behaviors. In addition to making substantial contributions to information and research on male fertility-related behaviors, this study has made two particular methodological contributions. First, it has demonstrated the feasibility of interviewing young men on these topics by obtaining good levels of response in both initial and followup interviews. Second, it has conducted an experimental assessment of audio-CASI methods for obtaining self-reports of sensitive behaviors. Initial findings indicate that audio-CASI methods increase self-reports of same-sex sexual behavior significantly over paper-and-pencil self-report methods. The most recent round of this study, conducted in 1995, included followup with the original panel interviewed in 1988 and 1991, as well as interviews with a new nationally representative sample of 1729 males age 15-19.
National Survey of Families and Households (NSFH). The NSFH is a national longitudinal survey addressing a broad range of topics related to family life. The first two rounds were conducted in 1987-88 and 1992-94. Within each of approximately 13,000 households, a primary respondent was selected and interviewed. The same interview was administered regardless of the respondent's gender. Much of the interview focused on children and parenting. Some couple data was obtained. Questions about the first husband/wife included whether he/she had been married before and/or had children at the time of the union. Union and birth transitions between waves is quite detailed, but there is limited information on nonresident unions. Both waves include a full range of relationship indicators for resident unions. Attitudes toward union formation and dissolution (both normative and personal) are included. Dating, sexual experience, and early family formation events are available for older focal children (age 13-18 in 1988, 18-23 in 1993), and the next younger group of focal children provide information on dating and sexual experience at the second wave.
National Survey of Family Growth (NSFG). The NSFG is a periodic survey of U.S. women ages 15-44 that has been conducted by the National Center for Health Statistics about every 5 or 6 years since 1973. The most recent cycle was conducted in 1995, and a public use data file has been released. The survey used innovative data collection techniques, including audio computer assisted self-interviewing (ACASI) to obtain detailed information about the respondent's life. The NSFG also asked about the men in the respondent's life. The methodological importance of this study is, in part, that it found that both incentives and the ACASI technique increased reporting of sensitive events and behaviors (e.g. abortions). In this application, the incentives promoted higher response rates and more than paid for themselves, evidently by creating a reciprocation between the respondent and the survey administrator. In addition, test interviews and expert analyses in the NCHS Questionnaire Design Research Laboratory and by Research Triangle Institute, the survey contractor, were instrumental in developing the NSFG life history calendar and procedures, as well as resolving many other questionnaire issues (Peterson and Schechter, 1995).
National Survey of Men. Although this 1991 study of sexual behavior and condom use among 20-39 year old men in the United States did not have a strong focus on fertility issues, it provides rare data on adult males' reproductive behaviors and sexual relationships. Thus, for up to 8 non-marital relationships that lasted 30 days or more since January 1990, the study collected information on pregnancies that occurred within each relationship, and the planning status and outcome of each (up to 3). It also collected information about the partners' demographic characteristics, and about sexual and contraceptive behavior in the relationship. The study cast a wide net in looking at relationships, including nonsexual relationships, nonmarital sexual relationships, and marriages and cohabitations. Some studies are underway using these data, and they may provide a valuable resource for understanding links between relationship characteristics and fertility risk. These data are unique because they focus on an older population of men that has received insufficient study in the past.
Panel Study of Income Dynamics (PSID). The PSID is a longitudinal study following an initial cohort of 5,000 families and their offspring since 1968. In 1997, the PSID will administer a Parent-Child Supplement, to include approximately 3,200 children under age 12. Respondents will include up to two children from about 1900 households, the primary caregiver of each child (e.g., biological, adoptive, step, or foster mother), the other caregiver of each child (e.g., the spouse of the primary caregiver or grandmother of the child), absent fathers, elementary or middle school teachers, preschool or day care teachers, in-home day care provide's, elementary or middle school administrators, and preschool or day care center administrators. Priority rules have been developed for defining order of inclusion in each of these categories. The different respondents will provide information through assessments, time diaries, and questionnaire booklets about the child and the household. If the biological father lives outside the household, the PSID will attempt to interview him, although it is not yet clear how difficult it will be to locate the absent fathers. In any case, the sample of absent fathers is likely to be small. The primary caregiver is also to be asked a battery of questions about the child's involvement with the absent father, so some data will be available from this perspective.
Survey of Income and Program Participation (SIPP). SIPP is a longitudinal household panel study conducted by the Bureau of the Census, with short- and long-term longitudinal components. It includes modules on child well-being, child care, child support, as well as information on income contributions and recipiency within the household and both to and from non-household members.
Surveys Conducted by the Bureau of Justice Statistics (BJS). BJS sponsors numerous surveys of inmates in jails, prisons, halfway houses, or probation agencies. Surveys such as the Survey of Adults on Probation (SAP), the Survey of Inmates in State Correctional Facilities (SISCF), the Survey of Inmates in Federal Correctional Facilities (SIFCF), and the Survey of Inmates in Local Jails (SILJ) generally achieve high response rates (with the exception of the SAP). Although most of the content is focused on criminal justice issues, survey items also include basic demographics, parental characteristics, questions about alcohol and drug use, and similar topics. More than ninety percent of inmates are male, and the great majority of these men are fathers. Questions about father involvement ask about the children's living situation before incarceration and currently, contact with children, and sources of economic support.
Methodological Issues
This section provides a review of several methodological issues related to gathering information on fathers and fatherhood. These issues have differential impact on studies of varying design, and so they are not insurmountable or uniformly challenging in all the studies that may be recommended for examining fathering. The methodological issues are loosely clustered into three groups: population identification, data collection procedures, and study designs. The impact of these issues for different types of studies is explored in more detail in the following section.
Population Identification:
Undercount. Fathers who are not located or are not included in the survey process at all are undercounted in large scale sample surveys. This includes the traditional undercount by the Decennial Census that affects the coverage of the sampling frame. Undercount rates are higher for men than for women, and for minorities than for whites and Asians. The undercount varies by age and race combined, ranging from 7 to 17 percent for black men. It is also related to household structure and relationships. Undercount rates are higher for unrelated persons, such as roomers, roommates, and men who are not married to the household respondent. It also appears to be greater for never-married fathers than for previously -married fathers. In addition, men in the military, prisons, jails, or other institutions are typically excluded from household surveys.
One promising technique for reducing the undercount in household surveys is to use expanded rosters with multiple probes. For example, the Census Bureau undertook an experimental "Living Situation Survey" in 1993 (Sweet, 1994) in which it oversampled minorities and renters, two sources of the undercount of fathers. The household roster section included a battery of roster probes. The first question was, "Who stayed here last night?" Another 3 percent of usual residents were elicited by the question, "Who lives here but didn't stay here last night?" For occasional rather than usual residents, a useful probe was, "Since [reference date], who lived or stayed here for one or more nights?" The survey identified an average of 1 additional person per household, and the gains were particularly large for black and Hispanic males age 18 to 29. Cantor and Edwards (1992) also used a similar list in experimental rosters trying to reduce within-household undercoverage in SIPP. (see Appendix L)
Other studies are planning dual rosters. As noted, the NLSY97 will include a household roster and a second roster of relevant individuals who live elsewhere, such as non-custodial parents, non-resident children, and so forth.
In future studies, it might be useful to develop a typology of living arrangements. Not only would this help with the creation of a list of terms and probes, but it also would move survey researchers beyond thinking in terms of traditional families. Work by anthropologists, such as Ruth McKay (McKay, 1993), would be useful here (Martin and de la Puente, 1993). Particularly important would be estimating the proportion of households falling into each category. This information would help in designing samples. Not requiring full names on rosters might improve coverage (Kearny, Tourangeau, Shapiro, and Ernst, 1993). Another technique which could be used in a limited way is network analysis. It is a useful way to explore extended families and/or complicated living or economic dependency arrangements (Knoke and Kuklinski, 1982). Unobtrusive observations of living patterns also might be useful.
The use of administrative records will help reduce not only undercoverage but also undercounting. Household members not identified by respondents can be found through these records. Absent family members, especially those institutionalized or homeless, also could be identified. Matches to Census records, already being done by Census and BLS, may be another way of estimating the number and types of people missed in our surveys. This will provide some estimate of the magnitude of the problem relative to the population as a whole (Couper and Singer, 1996). Different administrative lists can be used in conjunction with area frames in constructing multiple frame designs (Groves, 1989).
Unit nonresponse, especially in cross-sectional surveys, can result in both undercoverage and undercounting. The number and characteristics of household members, including absent parents and children, will not be known. To the extent these households are not missing at random, estimates of counts will suffer.
The interviewer's role in undercoverage and undercounting should be addressed. Vacancy checks could be conducted not only to find missing households, but also to evaluate interviewer reports (Clark, Kennedy, and Wysocki, 1993). The eligibility rates (both in terms of households and persons) obtained by individual interviewers could be compared to one another or to historical estimates. Techniques for persuading reluctant households should be explored, including ways for interviewers to introduce themselves and the survey to respondents (Groves, 1989). If the interviewer is effective at representing himself or herself and the survey, it will go a long way toward reducing the suspicions or concerns of reluctant respondents. In addition, the effects of type of nonresponse, noncontact versus refusal, need more study (Groves, Cialdini, and Cooper, 1992).
One way to reduce the effects of undercoverage and undercounting is weighting adjustment; however, this assumes a model which is not sensitive to nonignorable nonresponse (Raghunathan, Groves, and Couper, 1996). Not only do these models incorporate demographic information based on geography, but they also take into account the type of nonresponse. This work and other work being done jointly by Census and BLS also consider another area for research-- the effects of different patterns of longitudinal nonresponse.
Underreporting. Absent male parents tend to underreport their parental status to a large extent even though they are included in the survey interviews. In one survey, the National Survey of Families and Households, this accounts for more than half of the missing fathers (McLanahan and Garfinkel, 1996). Disparities between the number of women with previously-disrupted marriages who have children from those marriages at home and the number of men with previously-disrupted marriages with children living elsewhere are often great (Cherlin, Griffith, and McCarthy, 1980). Some studies have found the shortage of non-resident fathers to be largely confined to African Americans, though the factors contributing to this shortage include institutionalization (27 percent), undercount (53 percent), and underreporting (20 percent) (Sorensen, 1996). Beyond underreporting of fatherhood itself, there is also an issue of misreporting child support payments. In unmatched samples, it appears that fathers are much more likely to report giving child support than mothers are to report receiving it (Seltzer and Brandreth, 1994). But in matched samples, when both parents knew the sample was matched, reports were sometimes similar, but not always so (Braver, Fitzpatrick and Bay, 1991; Braver, Wolchik, Sandler, Fogas, and Zvertina, 1991; Smock and Manning, 1996; Sonenstein and Calhourn, 1990). In most surveys, the percentage of non-resident fathers who report providing support is substantially greater than the percentage of resident parents who report receiving it. The extent to which the undercount of adult males interacts with this problem is not known; some researchers presume that fathers who are included in studies are more likely to be paying support, leading to an overstatement of the frequency of providing support. (see Appendix K)
Underreporting of children and of other sensitive behavior may be reduced through technological advances in survey research. For example, the ACASI technology mentioned earlier has boosted reports of abortion in tests of women conducted by the National Center for Health Statistics; and so did paying a modest incentive (Mosher, Pratt, and Duffer, 1994). The ACASI interview involves giving the respondent the interviewer's laptop and a set of earphones. The respondent hears questions on the earphones which also appear on the screen. She or he then answers the questions by pressing a key on the laptop, so that the interviewer cannot hear or see what she is doing. In a pretest in 1993, 14 percent of women who received neither a payment nor the ACASI interview reported an abortion. Twenty-two percent of those who received a $20 payment but no ACASI reported an abortion. Twenty percent of those who received the 10 minute ACASI interview but no payment reported an abortion. And 30 percent of those who received both the $20 payment and the ACASI interview reported an abortion. Technology such as this should be tested and developed further for men.
Changing Family Structures. To date, most large scale sample surveys have reflected more traditional family models with parents living in marital situations within the same households or parents living singly. It has been less common for surveys to take into account multiple family forms, including cohabiting, unmarried couples; single parent families with nonresident, never-married fathers; families with other relatives playing important parenting roles in children's lives; and families with extended networks beyond households. During the life cycle of a family, the family type may well change with important consequences for the children. Current means of collecting information on family structure and relationships between family members, fathers outside households, and family networks are inadequate to help researchers and policymakers understand the complexity of fathering roles as they have evolved.
In multi-family households, CAPI methodology allows for creating spinoff cases with new family rosters, and this is now used in CPS. The same technique is available in CATI (Tucker, Casady, and Lepowski, 1991), although it can be more cumbersome. Spinoff cases could be created for parents or children not living in the household. These people would be linked to the household by special relationship codes in the original roster. Spinoff cases might also be used in longitudinal surveys to follow movers, similar to what is being done in SIPP.
Research to develop or improve any of these procedures will require large and/or targeted samples. Either census block or tract data might be used, but a more efficient method would be administrative records. The use of administrative records, however, raises issues of confidentiality and privacy.
Sampling Strategies. Although research on fathers and fatherhood should focus on all fathers, researchers and policy makers are interested as well in subsets of fathers. Frequently interest is focused on men who are relatively rare in the population, even though they are of increasing interest and may even be increasingly common. This would include, as examples, absent fathers in different subpopulations (e.g., by race or by age of children), fathers in different employment statuses, or stepfathers. They may be "rare" because they are a small percentage of fathers, or because fathers may exist in a particular status for only a relatively short time in their own or in the lifetime of their family. Problems of adequate sample size are exacerbated in analyses that need to cross-classify by race/ethnicity, socioeconomic status, age and gender of the child, or various configurations of families. Sampling strategies that sample children or sample parents, and that include institutionalized populations have different strengths and weaknesses. If children are sampled directly, coverage of children should be improved with concomitant reductions in coverage bias. Since most children live in households, there is nearly always an adult who can report on the child's contact with present and absent parents. Even proxy reports would provide at least minimal information. However, sampling parents may lead to higher rates of successfully locating and interviewing absent fathers directly, without relying on obtaining locating information in the child's household. Direct interviews with absent fathers could reduce bias in reports of certain types of information, although it is not yet fully established what types of information are most subject to such biases. Combined sampling approaches may hold the most promise for in-depth studies of parenting, although the ramifications of these for study design have not yet been fully explored.
One of the basic problems is the large sample size needed to arrive at an eligible sample which can provide enough statistical power. Either this will require money or the ability to piggyback on other research or find other cost effective approaches. In the case of a large, dedicated sample, mode of administration will be an issue, and it is unlikely that a personal visit will be practical. A telephone survey will not include those without telephones, unless a dual-frame design is used (Groves and Lepkowski, 1985). A mail survey would be difficult to administer, and the response rate would be low. Research which investigates the cost and error implications of the choice of mode would be useful (Groves, 1989). More efficient telephone sample designs have been developed in recent years which take advantage of list-assisted methods and matching to census public use files and administrative records, and these should be explored (Mohadjer, 1988). These new designs would be particularly useful for the targeting of specific subpopulations to reduce the cost of screening for eligibility.
The alternative of using ongoing surveys also may be attractive. Since these surveys vary according to mode and sample design, they could be used for different purposes. The new NLSY design will screen households for children and identify not only present but also absent parents. The National Immunization Survey has hundreds of thousands of screened numbers with some information about the households found. Many of these households have not been burdened with long surveys (Abt Associates, Inc., 1994). It is possible that a CPS supplement could be used for gathering information on fathers and linked to the other CPS data for the households. If the American Community Survey goes into production, it might serve as a data collection vehicle on a periodic basis.
Another problem which must be faced is the following of movers in longitudinal surveys, and such an operation will be important for measuring long-term outcomes. Much can probably be learned from the NLSY, SIPP, and other surveys which attempt to track respondents across significant periods of time. For example, SIPP has issued a memorandum detailing the most effective tracking techniques (Allen, 1994). However, they have not exclusively focused on fathers, a group which might present a particularly difficult challenge. Again, administrative records might be explored as a way of following families that separate. Finally, there are a number of weighting issues to consider. How are families which split apart weighted, and attrition in the longitudinal surveys will require using methods for censored data (Little and Rubin, 1987; Wiley and Sons; and Amemiya, 1985).
Within household sampling is of some concern if more than one child is involved or there are children with more than one father. The actual selection may not be difficult, but issues might arise if the person selected does not actually live in the household or is uncooperative compared to others in the household. Furthermore, a parent actually could have children from different generations, and the relationships may be very different.
Institutional Populations. Typically, large scale national surveys of the population are of the civilian, noninstitutionalized population only. Because of the particular issues being addressed in the search for improved information on fathers, it is clear that a large share of men excluded by these approaches are fathers (Harlowe, 1996). To fully understand the roles that men play in their children's lives -- and the types of influence they may have intentionally or otherwise -- it is important to examine better ways to obtain information from men in institutions and in the military population.
Data Collection Procedures:
Response Burden. Collecting information on or from fathers clearly increases interview time. There is a strategic issue as to which surveys should be affected and how, since response burden is a substantial issue in many large scale surveys that affects both their feasibility (from a financial and operational perspective) and the quality of the information provided when respondents tire. This problem very quickly reverts to what information is required and what are the best ways to obtain it, but it is also a fundamentally methodological issue regarding how to balance subject matter among the most appropriate, effective, and efficient surveys of differing designs and content.
There are two types of respondent burdens to face, but it is unclear how these will play out in terms of surveys of fathers. Furthermore, both types are affected by mode. The first is the burden associated with the difficulty of the task. This would include the length of the questionnaire, how many respondents are interviewed, and how difficult the questions are to answer (Groves, 1989; Tucker, Casady, and Lepowski, 1991; Schuman and Presser, 1981; Heberlein and Baumgartner, 1978; Herzog and Bachman, 1981; Sudman and Ferber, 1974; Silberstein, 1993; Dillman, Brown, Carlson, Mason, Saltiel, and Sangster, 1995; Herriot, 1977, Hermann, 1993). What effects these factors have on data quality will depend on the mode of administration (deleeuw and van der Zouwen, 1988; Groves, et. al.; Tucker, Casdy, Lepowski, 1991; Groves, 1989; Wiley and Hochstim, 1967; Rogers, 1976; Warriner, 1991; Sudman and Bradburn, 1973; Morgenstern and Barrett, 1974; Krosnick and Alwin, 1987; Miller and Dowenes-Le Guin, 1989; Conrad, Brown, and Cashman, 1993; Silberstein, 1989; Mullin, Cashman, and Straub, 1996; Hermann, McEvoy, Hertzog, Hertel, and Johnson). For instance, surveys done in person have the potential to be more burdensome because they can be longer and involve more complicated tasks.
In this case, burden may be tied closely to the extent of recall required, and recall has been a subject of intensive study. The saliency of experiences will be related to the ability to recall them, but the way the questions or memory probes are ordered and formatted also can matter (Schuman and Presser, 1981). The difficulty of the task also is affected by whether data collection will recur (Kaspryzk, Duncan, Kalton, and Singh, 1989). Recurrent data collection can be quite burdensome. If the collection is done too often, the respondent is likely to become annoyed. Infrequent collection might avoid this problem, but it can make the recall task more difficult and recontact will be more problematic. The more infrequent the contact, the longer the survey might become.
A considerable amount of research is needed to develop less burdensome data collection instruments for fathers and children. This would include the level of difficulty associated with different questionnaire formats under various modes. Research should be done on the problems associated with recall of family history and the usefulness of available records in the household. The feasibility of inserting and removing modules of questions in both cross-sectional and longitudinal surveys should be examined. The optimal frequency of data collection for recurring surveys should be determined.
The other concern is the burden accompanying sensitive items (Colombotos, 1965). Questions about income, sexual practices, drug use, and some health conditions can be very sensitive to some respondents. In addition, information about family relationships, critical in this case, is often difficult to obtain from respondents. Mode of administration is important here also in that distance from the interviewer can affect the respondent's feelings of privacy and confidentiality. Methods of reducing the burden associated with sensitive items have been investigated. These include randomized response techniques, (Groves, 1989) self-administered survey instruments, (Turner, Forsyth, Reilly, and Miller, 1996) and question order (Groves, 1989).
Reporting. For understanding different aspects of fatherhood and fathering, it may be more desirable to use proxy- or self-reporting. The trade-offs between the two fundamental ways of obtaining information are related to cost, accuracy, reliability, and accessibility to the respondent. While in some cases proxy responses provide entirely adequate information, in others information can only be obtained directly from the father who is being studied. Further research is needed on which areas previous partners or children are able to serve as proxy respondents and which ones require the additional expense of locating and interviewing the fathers to achieve the needed accuracy and reliability. When fathers must be contacted directly, there may be serious problems with accessibility of the respondent, so that targeted studies may be designed to gather information on a more limited sample.
The central question to ask about proxy response is whether it is less accurate than self response. It seems that it should be in most cases (Jones, Nisbett, 1972, Lord, 1980). However, empirical work, which is difficult to do, has shown this to not always be true. Some have speculated the relationships among family members will have an effect (Groves, 1989; Mathiowetz and Groves, 1989; Moore, 1988). The other reason results on this question might vary is that the accuracy of proxy reports could depend on the subject of the inquiry, the questionnaire strategies used to obtain the reports, or whether the proxy has first-hand experience concerning the information being sought (Miller and Tucker, 1993; Tucker and Miller, 1993; Kojetin and Miller, 1993; Cash and Moss, 1972; Kojetin and Mullin, 1995; Mullin and Tonn, 1993; Bickart, Blair, and Menon, 1994; Schwarz and Sudman; Menon, Bickart, Sudman, and Blair, 1995; Kojetin, Burnbauer, and Mullin, 1995; Kojetin and Jerstad, 1997). Nevertheless, in the case of men's reports of their children living elsewhere, it seems clear that there is indeed underreporting.
Administrative Records. For targeted topics, it may be feasible to obtain some information from administrative records. These can be linked to sample survey data to yield some more specific estimates. However, the usefulness of administrative records is highly dependent on the topics being studied and the availability of information in different records systems. In any given application, researchers must investigate whether access to records can be obtained under the auspices of the study, what information is available, the quality of the information in the system (primarily in terms of accuracy and completeness), and how such information might be linked to other data being obtained in the study.
Mode of Data Collection. The consequences of gathering data using different modes (mail, telephone, or personal interviews; degree of computer-assistance; observational studies; diaries; or other modes) are closely related to the type of study being undertaken. However, there is still considerable latitude in the designs of some research. Most studies of the effects of interviewing mode have been made with the more typical respondent--the mother or the child. Consequently, further research is needed into how these modes may influence data quality and response rates.
Study Design:
Questionnaire Design and Measurement Issues. There are a variety of issues related to the quality of information obtained from mothers and fathers about the role of fathers in children's lives. If either or both parents are interviewed, most surveys currently ask them both the same questions. Researchers are not yet certain what to ask fathers, because studies have not yet pointed to any distinctive understanding of fathers' roles. However, since researchers do acknowledge that fathers may have unique ways of interacting with their children, it is clear that such relationships cannot be discerned using traditional questions. Further research is needed on what aspects of fathering are important to men, what aspects of fathering are important to children, and ways to improve the quality of information collected. Specifically, the stability, reliability, and validity of survey responses are likely to be increased by improving the questions asked. Another measurement of critical importance is the time reference used in sample surveys. This also has important implications for the quality of data obtained from respondents.
New questions will be needed to assess what fathers contribute to their children, both emotionally and physically. Other questions will focus on the ways fathers and children view their relationships with one another. Some questions will be subjective, but many should be behavioral measures. Types of questions which could be used are attitude scales, behavioral frequency measures, behavioral checklists, and open-ended items (Poister, 1978).
Whatever the types of questions used, they must be thoroughly tested. This research is important to ensuring ultimately data quality. The identification of question wording and order effects is becoming commonplace, and the methods used in this area are growing (Groves, 1989; Schwarz and Sudman, 1993; Conrad and Brown, 1995; Esposito, Campanelli, Rothgeb, and Polivka, 1991; Forsyth, Lessler, and Hubbard, 1992; Turner, Lessler, and Gfroerer,; Martin and Polivka, 1995; Menon, 1994, Schwarz and Sudman; Willis, Royston, and Bercini, 1991; Tanur, 1992; Sudman, Bradburn, and Schwarz, 1996).
This work also examines problems of respondent understanding, memory, and recall, which will be of central importance in the development of data collection instruments concerning fatherhood. Small field and laboratory tests will be necessary, as well as the field observation of large-scale tests. Testing will involve think-aloud interviews, respondent and interviewer debriefings, and interview monitoring with behavior coding. Administrative data can also be used to measure data quality (Moore and Marquis, 1989).
Research should be undertaken to develop methods which overcome problems of memory and recall. Some research has already been done in this area (Anderson and Conway, 1993; Schwarz and Sudman, 1993; Burt, Mitchell, Raggatt, Jones, and Cowna, 1995), but more is needed as it relates to the experiences of fathers and children. One method which could have some merit is time-use diaries (Juster and Stafford, 1985). Respondents also might be asked to do narrative histories of family relationships which could be content analyzed (Dillman, 1978; Groves, 1989; Groves and Kahn, 1979). Other qualitative methods will be discussed under recommendations number 7, below.
Questionnaire design is dependent on the mode of data collection. For instance, long lists requiring flashcards cannot be used in telephone surveys, and lengthy narratives cannot be collected over the telephone. Question order effects will differ by mode, and the ability to obtain answers from multiple household members will be limited with both telephone and mail surveys. Literacy is a problem in mail surveys, but privacy and confidentiality is better preserved, unless computerized self-administered surveys are used. Thus, the effects of mode on surveys of fathers will need to be considered, and the information to be collected should be fitted to the mode.
Finally, multiple measures from multiple sources will be needed to ensure the quality and/or accuracy of the data. This is true for two reasons. As with most social research, the measures used can have a considerable amount of nonsampling error, so it is better to use multiple measures of the same concept and arrive at a combined indicator by "triangulation," also known as the multi-trait/multi-method approach (Alwin, 1974; Campbell and Fiske, 1959; Jick, 1979; Tucker, 1992). The other reason it is important, especially in this case, is that different respondents may give conflicting answers or, at least, express different points of view. A more accurate picture is likely to be obtained by asking the same questions to several family members and/or gathering data from outside sources such as education or health providers and administrative records.
Linking Quantitative and Qualitative Designs. Enhancing quantitative survey designs with qualitative research methods has the potential to enhance knowledge in at least two ways. First, researchers can address many of the methodological and substantive issues that are not yet completely or even well understood using qualitative techniques. Such smaller scale studies frequently provide approaches to address issues in large scale quantitative studies. These studies can help to develop topics to study, question wording, or survey design, as a few examples. Secondly, linking methods can greatly enrich what can be learned by either approach taken alone. Combined approaches provide a much more rounded view of social phenomena by calling on the strengths of each (U.S. Department of Education, National Center for Education Statistics).
The fact is that quantitative data, especially when presented only at the aggregate level, often masks or even misstates important relationships (Copeland and White, 1991). Thus, qualitative methods are needed to inform and guide quantitative research. Fortunately, the last decade has seen one example of the effective use of both methodologies--the use of cognitive methods in survey research (Fienberg and Tanur, 1989; Forsyth, Lessler, and Hubbard, 1992; Nargundkar and Gower, 1991; Turner, Lessler, Gfoerer; and Tanur, 1992; Sudman, Bradburn, and Schwarz, 1996). This combination was used very effectively in both designing and analyzing the Supplement on Race and Ethnicity to the Current Population Survey (Cannell, Oksenberg, Fowler, Kalton, and Bischoping, 1989; McKay and de la Puente, 1995; McKay, Stinson, de la Puente, and Kojetin, 1996; Tucker, 1996; Esposito, Campanelli, Rothgeb, and Polivka, 1991; Willis, Royston, and Bercini, 1991; Conrad, and Brown, 1995; Peterson and Schechter, 1995) Qualitative methodologies from other fields, such as anthropology, also have been used (McKay, 1993), and work is ongoing to include other disciplines like linguistics. (1)
Many lessons have been learned from these experiences. One of them is that qualitative methods are useful for designing questionnaires that interviewers can administer more easily and that respondents can understand. These techniques also can help explain seemingly conflicting or confusing findings from quantitative research. Several limitations, however, have already been encountered, and research is beginning to deal with these. One problem with qualitative research is that the methodology is less codified. This problem has been examined (Tucker, 1996), and more rigorous methods are being developed (Conrad and Blair, 1996; Tucker, 1996; King, Keohane, and Verba, 1994; Yin, 1989). Basically, qualitative research still must be judged against the same scientific standards as quantitative research. At the same time, qualitative research should not be dismissed out of hand if its standards are high. Therefore, in studies of fathers, the two should be used together, and the results should be judged with the same ruler. Given, the complexity of the problem, both will be useful.
Longitudinal or Cross-sectional Designs. Whether a longitudinal or cross-sectional design is selected is dependent on the kinds of information that are being sought. While longitudinal designs tend to be thought of as more expensive, they may be more cost-effective through providing richer information with a smaller sample than may be achieved with repeated cross-sectional studies.
Population Diversity. Just as it may be inadequate to study parenting by asking the same questions of both mothers and fathers, it also may lead to inadequate understanding of important issues if studies do not account for the diversity in the population. The rich cultural, ethnic, racial, and linguistic diversity in the population of the United States means that studies have to be carefully designed to elicit information from different groups. In studies that are characterized by uniformity of administration to all respondents (such as large scale sample survey research), this means that conscious compromises will need to be made to develop items that are understandable to a wide variety of respondents. In other types of research, special, more targeted, approaches may be taken when dealing with different populations; or specific studies may be developed for different groups. The challenges of population diversity relate to the content of the study (different aspects of fathering may carry different levels of importance), conceptualization of the content (different groups may have varying perspectives on the same issues), and structure and wording of questionnaires or interview templates.
Measuring Time Use. Assessment of parent-child interaction often rests largely on reports of children's time use. There are several ways of assessing how much time and in what activities parents and children engage. The most accurate way to collect such data would be through observation. However, such methods are costly, intrusive, and limited in the amount of a day that can be covered. Another accurate way to collect information is by time sampling, in which respondents write down the activity they are engaged in whenever a beeper sounds. This methodology is also costly, intrusive and limited. The most common method in survey research is to ask parents directly how much time they spend in certain activities, such as reading to their child. While simple and widely used, this method is known to be biased. First, it is subject to social desirability bias. Parents will report more time spent on desirable activities (such as reading) than on less desirable ones. Second, there is no baseline against which to check consistency, validity, or reliability. Thus times have been shown to be quite inaccurately reported (Juster and Stafford, 1985).
In contrast, substantial methodological work has established the validity and reliability of data collected in time-diary form (Juster and Stafford, 1985). The instrument for assessing time use is a "time diary," which is a chronological report by the child and/or the child's primary caretaker about the child's activities over a specified recent 24-hour period, beginning at midnight (who the reporter is depends on the age of the child). The time diary is interviewer-administered and asks several questions about the child's flow of activities, such as what they were doing at that time, when the activity began and ended, and what else they were doing (if they were engaged in multiple activities). The Child Development Supplement to the Panel Study of Income Dynamics also added two questions: "Who was (child) doing that with?" And "Who else was there?" These added questions, when linked to activity codes such as "playing" or "being read to" provide unbiased details on the extent of parent/child one-on-one interactions and availability of the parents. The advantage of such questions is that total time in one day has to add to 24 hours. Consequently, while individual times may be slightly inaccurate, the times are consistent with one another. The disadvantage of the time diary is that it represents only a sample of children's days. Thus while it accurately represents the activities of a sample of children on a given day, it is only a very small sample of a given child's days and, as such has limited reliability. To improve reliability, most time-use studies obtain at least one weekend and one weekday assessment, and many also obtain multiple samples over a period of time, such as a year.
Since the data collection format is open-ended--an advantage for avoiding biases toward "good" activities and away from "bad" activities but a potential pitfall for proper interpretation of the data--precise, clear, and well-focused definitions of activities are vital. Fortunately, the 1975-1981 Time Use Study has paved the way in terms of guidelines for coding children's time-diary reports (Juster and Stafford, 1985). Working with several child development experts and time-use experts in a number of disciplines and representing a wide range of countries and cultures, Hill, Stafford, Juster, and colleagues in the 1981 follow-up in the 1975-1981 Time Use Study spent considerable time and effort designing a time-use methodology appropriate to children (Hill, Stafford, Juster, and colleagues, 1975-1981). The methodology is not onerous. Researchers have found that parents of young children enjoy working with their youngsters to provide the children's time diaries, which take about 15 minutes per child per day, and can adequately represent the child's day.
How Should New Data Collection Be Undertaken?
There are two issues that are fundamental for the research community to consider in designing studies to obtain information on fathers: (1) Should a new study be initiated or would an add-on to an existing study be more appropriate? (2) Should the study be conducted by Federal statistical agencies or as a privately sponsored effort? While it is clear that the right directions depend in part on the nature of the study, some guidance about factors to consider when addressing these questions may be useful.
New vs. Supplemental Studies. In the past, new ideas may have readily generated entirely new studies. However, concerns about financial support for social science research now more often lead to consideration of ways to piggyback onto existing studies. There are, however, advantages and disadvantages to either approach. New studies have a distinct advantage in that the designers and sponsors of the study can exercise substantially greater latitude in defining the scope of the study. As a result, they are better able to focus the entire study on the topics of interest rather than having to fit components in around an existing questionnaire or other information collection. Similarly, they have greater control over the research design and study operations, within cost constraints, so that these aspects can be tailored to their needs. They have the disadvantages of higher costs and longer start-up time that unavoidably occur with a new program.
Supplemental studies address that disadvantage directly, typically being of lower cost and with a faster start-up time. Often, the sponsor will only have to contribute marginal costs, which may be minimal, to obtain additional information. Additional information that is likely to be related to the topic of interest will be obtained at no cost to the sponsor because it is included in the base survey. However, the lack of control over the design of the survey and the sample introduce potentially severe disadvantages. The sponsor may not have control over question wording, although this problem is more likely to affect those items already in the study than those being added. Lack of control over survey operations and data processing can hinder the utility of the results, insofar as they influence the outcomes of the inquiry or the timeliness of reporting. The latter is a problem particularly if the primary data are processed with higher priority. In a related issue, the context of the independent study may introduce response or nonresponse bias if its content or design are not compatible with the goals of the sponsor. Finally, although the cost advantage is attractive, this approach means that the sponsor depends on another organization to carry out the survey, to obtain funding for the core, and to produce the data. It is not entirely unusual for such arrangements to fall through when funding unfortunately becomes unavailable for the sponsor of the core survey.
Federal vs. Privately Sponsored Studies. Studies that are conducted by or for the Federal government under contract have different strengths and weaknesses than those of studies that are fully privately sponsored or that are conducted under a grant from the Federal government. However, in recent years, these distinctions have become increasingly blurred, as funding sources for Federal statistical studies have declined and the quality of large scale research in the private sector has improved. Nevertheless, there remain significant differences between these two types of studies.
While Federal studies historically have been thought to have more secure funding sources once the government committed to the survey, this may no longer be the case in the current budget climate. Federal surveys do have a small advantage in easier access to national sampling frames that may be more difficult to construct in the private sector. Federal researchers are constrained to create public use analysis files for researchers to have equal access to, and this clearly enhances the value of the study for the broad research and policy community. Federal agencies typically provide metadata, describing the characteristics of the data, which is highly important for more informed use of the data. Consistent with these last two aspects, publicity about the availability of the data is typically seen as a part of the survey process, thus enhancing access. Despite concerns about response burden, the Federal government still tends to achieve substantially higher response rates than are achieved in privately-sponsored surveys.
Federal surveys also have disadvantages, many of which stem from a generally long lead time from conceptualization to development to data production and analysis. Funding is typically difficult to secure initially. The clearance process conducted by the Office of Management and Budget adds considerable time to the survey process, and can place constraints on the response burden and content of surveys that can restrict the topics that can successfully be studied. In certain studies, association of a study with a particular agency may introduce response bias. Finally, whether because of elaborate designs or operational inefficiencies, Federal studies tend to be somewhat costly.
Privately sponsored studies or studies conducted with Federal grants avoid some of these disadvantages. The researcher may have more latitude in defining the topics of study, and so may be able to address more sensitive issues. These studies show a clear advantage in that less time is typically required to move from conceptualization to data production. Researchers can more readily adopt innovative techniques, that may (or may not) prove useful from a wide variety of perspectives. And, as noted above, such studies may be designed to serve more precise needs and they may, as a result, to be less costly.
On the other hand, privately sponsored studies are less likely to provide timely public use data files to allow the broader research community access for analysis. The degree of collaboration is more dependent on the individual principal investigators, as there is less motivation to do so from the perspective of the public good. Finally, the care and attention focused on technical issues of all sorts varies considerably in such studies. This disadvantage can have serious and broad consequences for the quality and utility of the data.
Recommendations
Let us now summarize the implications of the research activity we have reviewed for future research on fathers. The state of knowledge about how to study fathers is not adequate to prescribe a single set of optimal procedures for all studies; and we do not wish to create a new methodological orthodoxy. Nevertheless, we believe that the following implications can be drawn.
1. Include fathers. Fatherhood is a complex aspect of our society that is inadequately understood. The knowledge base is insufficient to inform policy makers about the roles that fathers and mothers play in our families and our communities. Issues extend beyond the most commonly expressed concerns about absent fathers. Thus, national surveys need to provide a more accurate and in-depth profile of fathers to improve this understanding. Two surveys in particular should consider including fathers as interviewees - the Early Childhood Longitudinal Survey and the National Survey of Family Growth.
Studies of what non-resident fathers do should include non-resident fathers. Although this precept might seem self-evident, its adoption would mark a major change in research design. Until recently, an inordinate proportion of studies of fatherhood have attempted to measure the importance of absent fathers solely by examining households in which fathers are absent. In most of these studies, little or no effort was made to contact absent fathers. This literature on father absence has been useful but it has its limits. Studies that dichotomize all fathers into "present" and "absent" may miss important aspects of a child's continuing relationship with a non-resident parent. Studies that do not contact the absent parent are inherently limited in the understanding they can provide about why fathers may be relatively uninvolved with their children. Future research on non-resident fathers should move beyond merely studying their absence.
2. Improve household survey methodology. The standard household survey methodology is critical to our understanding of fathers because it is the only methodology that has the potential for identifying the entire universe of resident fathers and nearly all nonresident fathers. A very small share of fathers are outside of this sampling frame. Part of the underrepresentation of fathers in household surveys is due to an undercount of fathers who are tenuously attached to households and part is due to underreporting by men who are interviewed but who do not disclose that they have children living elsewhere. Both of these issues can and should be addressed.
3. Add expanded household and extra-household rosters to existing surveys. Standard rosters in household surveys are not adequate to resolve the problems of underrepresentation. Experimental surveys have increased their coverage of underrepresented groups of fathers by using an expanded set of questions and probes. Existing surveys should test these questions and probes along with their standard rostering techniques. Follow-up interviews should be conducted with a subset of these individuals to ascertain who is not being interviewed. Some surveys are also obtaining extra-household rosters of important family members who live elsewhere, such as non-resident parents and non-resident children and attempting to conduct follow-up interviews with these individuals. Further study of these individuals may be desirable.
In-depth studies (particularly long-term longitudinal studies) should carefully consider whether including fathers as interviewees would not improve the utility of the database. The Early Childhood Longitudinal Study of the National Center for Education Statistics, in particular, should make every effort to include a father supplement at some point in the study. This study, currently under development, could provide important information about children's development in relation to father involvement that could have important policy implications. Some effort is needed to include at least correctional institutions in household surveys to fill out the picture of absent fathers. The typical exclusion of men in institutions leads to a distorted view of how families function in our society.
4. Develop questions that are relevant to fathers and result in accurate responses. Unlike the well-tested interview protocols for female fertility and family formation, protocols for surveys focused on fathers are not yet well-developed or standardized. It is not wise to merely ask fathers a set of questions about parenting that parallels the set typically asked of mothers. Rather, new questions are needed to assess fathers' contributions to their children's development. Better measures of time use, such as time diaries, need to be incorporated in studies. Consequently, survey-based studies of fathers should include a substantial amount of development and testing prior to interviews with the sample. Exploratory methods exist that use laboratory and small field-test settings; these methods include think-aloud interviews, respondent and interviewer debriefings, and interview monitoring.
5. Improve procedures for asking sensitive questions. There is strong evidence that even when fathers are interviewed, they underreport the existence of their children living elsewhere. Mothers may also underreport non-resident fathers of their children. In addition, non-resident parents may be motivated to exaggerate the amount of contact they have with their children. For these reasons, it is important to employ, when feasible, improved measures for obtaining this information. We will note below that this is an important topic for further methodological research. Some promising techniques for survey research have been developed, such as audio computer self-administered segments of interviews. This is also a topic for which ethnographic studies are useful, both for identifying and studying fathers whose existence may not be revealed by a survey and for suggesting better ways to ask sensitive questions in surveys.
6. Reduce response burden. Other Working Groups recommend placing a high priority on obtaining detailed sexual, reproductive, and union histories for men. Yet the complexities of some men's sexual histories and reproductive careers means that for a subset of fathers, obtaining comprehensive histories could impose a substantial response burden. The very fathers who have the longest, most complex histories are often the group of greatest interest. It is not clear how much information can be collected from them: respondents may tire of remembering their histories at some point, or they may remember dates inconsistently. Therefore, a high priority for methodological research is to undertake studies of ways to reduce the response burden imposed by extensive histories. The life-history calendar is one way to reduce the burden; it seems to be clearly preferred by respondents to interviewer questions; and it seems to result in better quality data (Peterson and Schechter, 1995). But little methodological research has been conducted specifically on men. New studies that propose the collection of extensive sexual, reproductive, and union history from men should include development and pretesting of ways to reduce the response burden of histories; and methodological research on the topic should be supported.
7. Conduct intensive observational studies. The gaps in our knowledge of what fathers (both resident and non-resident) do suggest the importance of smaller, intensive observational studies. For example, developmental psychologists conduct studies of children and their caregivers that involve direct observation, batteries of tests and assessments, and sometimes videotaping and subsequent rating of family interaction. Ethnographers conduct studies that use anthropological field work methods to describe and understand family interaction. These kinds of studies can provide valuable insights about fathering. They also can serve as hypothesis-generating studies that yield propositions about fathers that can be tested by subsequent close-ended questions in larger, more representative sample surveys.
8. Use supplementary and alternative sampling strategies. The standard household sample-survey methodology appears not to find many unmarried fathers. Other sampling strategies may sometimes be advantageous, either as supplements to household samples or as alternatives to them. The underrepresentation is particularly large for young men from minority groups, so other sampling strategies are particularly important for studies which focus on them Part of the underrepresentation is due to an undercount of fathers who are tenuously attached to households and part is due to underreporting by men who are interviewed but who do not disclose that they have children living elsewhere. The other sampling strategies include the use of administrative records to locate fathers who may no longer be involved with their children or whose names are not supplied by respondents in a household survey. They also include the addition of the incarcerated population and the military population when possible. In addition, they may include the development of alternative designs such as sampling on births at hospitals and interviewing both parents for the first time as soon after the birth as possible.
9. Recognize population diversity. The roles of fathers are embedded in larger family processes that can differ by class, race, and ethnic groups. Even within these groups there can be substantial diversity. Studies need to take this diversity into account. For example, studies of low-income groups where single-parent families and broad kinship networks are more common should consider the roles in children's lives of stepfathers, male kin, mothers' boyfriends, and other men . In addition, the roles of biological fathers may differ in family settings where extended kinship ties (such as to grandparents, aunts, or uncles) are present; consequently, studies of fathers should consider variations in family patterns.
10. Be careful of unobserved sources of bias. Despite the best efforts of researchers, studies of fathers can suffer from bias due to incomplete observation or to patterns of responses to questions. Fathers who are underrepresented are likely to have some characteristics that differ from fathers who are represented. Data collected from mother-father pairs are, in principle, superior to data collected from only one parent; but, in practice, the difficulty of collecting matched mother-father reports can result in an underrepresentation of certain kinds of couples. Difficulties in attributing cause and effect can arise, particularly in cross-sectional studies. Studies of fathers should at the very least demonstrate that such problems, and their likely effects on analyses, have been considered. Research designs that can reduce bias should be used where possible. These include so-called panel data (longitudinal studies that can be used to control for unchanging unobserved sources of bias), studies of families that are affected by external assignments of fathers' roles such as military transfers or court orders, and statistical models that attempt to correct for incomplete and self-selected samples.
11. Carefully consider additions to existing data programs. It is not clear that completely new, large-scale studies should be undertaken to investigate issues related to fatherhood and fathering at this time. There is a great deal to be learned from working with existing survey mechanisms to expand the content and scope of studies in targeted, appropriate ways to address specific questions. Very little is understood on this topic to inform an emerging policy debate that encompasses far more than just the economic role of fathers. Consequently, important contributions can be made with small scale work and through expansions to existing studies of family conditions and processes. In this time of scarce resources for social science research, funds should be directed where they will provide the greatest insights. Thus, careful trade-offs need to be made in investing in new studies, major expansions of existing studies, and continuing some existing data collections as is in the interests of economy.
12. Conduct more methodological research. Lastly, we call for a program of methodological research on studying fathers. Because of the focus of past studies on mothers and on families that do not extend beyond the boundaries of one household, not enough is known about how to study fathers. We have briefly summarized the major developments in methodological research at this time. But many important facets of research on fathers need to be improved before we can be satisfied with the quality of current and future studies. These include the basic problems of finding non-resident fathers, of the underreporting of fatherhood among the men that are found, and of obtaining full and accurate answers about contact with children living elsewhere. Solutions involve sampling strategies, interviewing techniques, and questionnaire design. We need to know more about how to combine and analyze responses from mother and fathers (coresident or non-coresident) in data in which couples are the unit.
Furthermore, in order to construct informative surveys, we need to know more about what aspects of fathering are important and valuable. Questions of and about fathers should include more than just their economic circumstances and contributions to families. An expanded concept of fatherhood is essential. We doubt that this information can be obtained without detailed, observational studies of fathers and children of the type carried out by developmental psychologists and ethnographers. Technically speaking this is substantive, rather than methodological research; but it is a necessary precursor to the construction of adequate structured survey instruments. For example, other Working Groups note the importance of determining men's attitudes toward fatherhood and their motivations for having children. Although there is a psychometric literature on the reliability and validity of survey-based measures of attitudes and motivations, little research has been conducted on the population of interest. Qualitative studies would be particularly useful in order to determine the kinds of questions that close-ended surveys ought to ask.
Conclusion
There is much methodological work to be undertaken to help improve the quality and scope of information available on fathers and fathering. This paper has attempted to present some of the methodological issues, while at the same time suggesting some types of activities that could be undertaken immediately to improve the information base. Although it is always difficult to discuss methodology absent a clear concept of what content is needed, the dearth of information on these topics is so severe that some actions must be taken immediately. It is the sincere hope of this Working Group that the Fatherhood Conference will provide a strong foundation from both the substantive and methodological perspectives to support moving expeditiously to fill the data gaps.
References
Abt Associates, Inc., "The National Immunization Survey: 1994 Annual Methodology Report," (Still under review)
Allen, T. 1994. "SIPP: Results of the July and August mover questionnaires," Memorandum, Bureau of the Census.
Alwin, D. 1974. "Approaches to the interpretation of relationships in the multitrait-multimethod matrix." Sociological Methodology 1973-74. San Francisco: Josey-Bass
Amemiya, T. 1985. Advanced Econometrics, Cambridge: Harvard University Press
Anderson, S.J. and Conway, M.A. 1993. "Investigating the structure of autobiographical memory." Journal of Experimental Psychology: Learning, Memory, and Cognition 19:1-19
Babbie, E. 1989. The Practice of Social Research, Fifth Edition, Belmont, CA: Wadsworth.
Braver, S.H., Wolchik, S.A., Sandler, I.N., Fogas, B.S., and Zvetina, D. 1991. "Frequency of visitation by divorced fathers: Differences in reports by fathers and mothers." American Journal of Orthopsychiatry 61:448-454
Braver, S.L., Fitzpatrick, P.J., and Bay, R.C. 1991. "Noncustodial Parents' Report of Child Support Payments." Family Relations 40:180-185
Burt, C.D.B., Mitchell, D.A., Raggatt, P.T.F., Jones, C.A., and Cowan, T. M. 1995. "A snapshot of autobiographical memory retrieval characteristics," Applied Cognitive Psychology 9:61-74.
Campanelli, P. and Sturgis, P. 1996. "Collecting survey introductions: Initial results," Paper presented at the 7th International Workshop on Household Survey Non-Response, Rome, Italy.
Campbell, D.T. and Fiske, D.W. 1959. "Convergent and discriminant validation by the multitrait-multimethod matrix." Psychological Bulletin 56(2):81-105
Cannell, C. F., Fowler, F. J., and Marquis, K. H. 1965. "A report on respondents' reading of the brochure and letter and an analysis of respondents' level of information," Ann Arbor: Institute for social Research, University of Michigan
Cannell, C., Oksenberg, L., Fowler, F.J., Kalton, G., and Bischoping, K. 1989. "New Techniques for Pretesting Survey Questions," Final Report for Grant Number HS 05616 from the National Center for Health Services Research and Health Care Technology Assessment, Ann Arbor, MI: Survey Research Center, University of Michigan
Cantor, D. and Edwards, C. 1992, "Testing Alternative Household Roster Questions for the Survey of Income and Program Participation." Final Report submitted to the Census Bureau under Bureau of Labor Statistics contract #J-9-J-8-0083.
Casady, R.J. and Lepowski, J.M. 1993. "Stratified telephone survey designs," Survey Methodology 19(1):101-113.
Cash, W. S. and Moss, A. J. 1972. "Optimum recall period for reporting persons injured in motor vehicle accidents," Vital Health Statistics, National Center for Health Statistics, Series 2, No. 50
Cherlin, A., Griffith, J., and McCarthy, J. "A note on maritally-disrupted men's reports of child support in the june 1980 current population survey." Demography 20:385-389.
Cherlin, A.J. 1992. Marriage, Divorce, Remarriage. Revised and Enlarged Edition. Cambridge: Harvard University Press.
Clark, C., Kennedy, B., and Wysocki, M. 1993. "Coverage Error in the Canadian Labour Force Survey." Proceedings of the Bureau of Census Annual Research Conference. 602-637
Colombotos, J. 1965. "The effects of personal vs. Telephone interviews on socially acceptable responses," Public Opinion Quarterly 29(3):457-458
Conrad, F. and Brown, N. 1995. "Estimating Frequency: A Multiple Stragegy Perspective," Paper presented at the annual conference of the Association for Consumer Research, Minneapolis, MN
Conrad, F., Brown, N., and Cashman, E. 1993. "How the memorability of events affects frequency judgments," American Statistical Association Proceedings of the Section on Survey Research Methods 1058-1063
Conrad F. and Blair, J. 1996. "From Impressions to Data: Increasing the Objectivity of Cognitive Interviews," Paper presented at the annual meeting of the American Statistical Association, Chicago, IL
Copeland, A.P. and White, K. M. 1991. Studying Families, Newbury Park, CA: Sage Publications
Couper, M.P. and Singer, E. 1996. "Noncoverage error and the NCVS," Report submitted to the Bureau of Justice Statistics.
Couper, M. and Groves, R.M. 1992. "The role of the interviewer in survey participation." Survey Methodology 18:263-271
Dawes, R. and Smith, T. 1985. "Attitude and opinion measurement." In Lindsey, G. Handbook of Social Psychology, New York: Random House
deleeuw, E.D. and van der Zouwen, J. 1988. "Data quality in telephone and face-to-face surveys: "A comparative meta-analysis." In R. M. Groves, et. al., (Eds.), Telephone Survey Methodology. New York: Wiley, 283-300
Dillman, D. 1978. Mail And Telephone Surveys, New York: Wiley & Sons;
Dillman, D.A., Brown, T.L., Carlson, J.E., Mason, R., Saltiel, J. and Sangster, R.L. 1995. "Effects of Category Order on Answers to Mail and Telephone Surveys." Rural Sociology, 60(4):674-687
Esposito, J.L., Campanelli, P.C., Rothgeb, J.M., and Polivka, A.E. 1991. "Determining which questions are best: Methodologies for evaluating survey questions," American Statistical Association Proceedings of the Section on Survey Research Methods 46-53; b.
Esposito, J.L., Campanelli, P.C., Rothgeb, J.M., and Polivka, A.E. 1991. "Determining Which Questions are Best: Methodologies for Evaluating Survey Questions." American Statistical Association Proceedings of the Section on Survey Methods 46-53
Fienberg, S.E. and Tanur, J.M. 1989. "Combinging Cognitive and Statistical Approaches to Survey Design," Science 243:1017-1022
Forsyth, B.H., Lessler, J.T., and Hubbard, M.L. 1992. "Cognitive Evaluation of the Questionnaire." In C. F. Turner, J. T. Lessler, and J. C. Gfoerer (Eds.), Survey Measurements of Drug Use: Methodological Studies, DHHS Publication No. 92-1929, Washington, DC: US Government Printing Office
Forsyth, H. Lessler, J.T., and Hubbard, M.L. 1992. "Cognitive Evaulation of the Questionnaire." In C. F. Turner, J. T. Lessler, and J. C. Gfroerer (Eds.), Survey Measurements of Drug Use: Methodological Studies, DHHS Publication No. 92-1929, Washington, DC: US Government Printing Office
Groves, R.M., Cialdini, and Cooper, M.P. 1992. "Survey Participation." Public Opinion Quarterly 56:475-495.
Groves, R.M. 1989. Survey Errors and Survey Costs, New York: John Wiley. Ch. 3.
Groves, R.M. 1989. Survey Errors and Costs, New York: John Wiley, Ch. 5 and 11
Groves, R.M. 1989. Survey Errors and Survey Costs, New York: John Wiley. Ch. 10; and A.
Groves, R.M. 1989. Survey Errors and Survey Costs, New York: John Wiley. Ch. 9.
Groves, R.M. and Lepkowski, J.M. 1985. "Dual Frame, mixed mode survey designs." Journal of Official Statistics 1:264-286
Groves, R.M., and Kahn, R.L. 1979. Surveys by Telephone, New York: Academic Press
Harlowe, C. 1996. Communication with this Working Group, 1996.
Heberlein, T.A. and Baumgartner, R. 1978, "Factors affecting response rates to mailed questionnaires: A quantitative analysis of the published literature," American Sociological Review 43:447-462
Hermann, D.J. 1993. "The validity of retrospective reports as a function of the directness of retrieval." In N. Schwartz & S. Sudman (Eds.), Autobiographical Memory and the Validity of Retrospective Reports. New York: Springer-Verlag, pp.21-37.
Herriot, R.A. 1977. "Collecting Income Data on Sample Surveys: Evidence from Split-panel Studies." Journal of Marketing Research 14:322-329
Herzog, A.R. and Bachman, J. 1981. "Effects of questionnaire length on response quality," Public Opinion Quarterly 45:549-559
Hill, F. P. Stafford, F. T. Juster, and colleagues. 1975-1981 Time Use Study. 1981.
Hochstim, J.R. 1967. "A Critical Comparison of Three Strategies of Collecting Data from Households," Journal of the American Statistical Association, 62(319):976-989
Hughes, A.L. and Gfroerer, J.C. 1990. "Analysis of survey data on drug experience by mode of data collection," Proceedings of the Section on Survey Research Methods, Anaheim, CA
Jick, T.D. 1979. "Mixing qualitative data as an attractive nuisance: The problem of analysis." Administrative Science Quarterly, 24 (4):602-611
Johnson, T.P., Hougland, Jr., J.G., and Clayton, R.R. 1989. "Obtaining reports of sensitive behavior: Comparisons of substance use reports from telephone and face-to-face interviews," Social Science Quarterly, 20(2):174-183
Jones, E.E. and Nisbett, R.E. 1972. "The actor and the observer: Divergent perception of the causes of behavior." In E.E. Jones, et. al. (Eds.), Attribution: Perceiving the Causes of Behavior, Morristown, NJ: General Learning Press
Jones, E.E. and Nisbett, R.E. 1972. "The actor and the observer: Divergent perception of the causes of behavior." In E.E. Jones (Eds.), Attribution: Perceiving the Causes of Behavior Morristown, NJ: General Learning Press
Lord, C.G. 1980. "Schemas and images as memory aids: Two modes of processing social information," Journal of Personality and Social Psychology, 38(2), 257-269.
Juster, F.T and Stafford, F.P., (Eds). 1985. Time, Goods, and Well-Being, Ann Arbor, MI: Institute for Social Research; and Social Indicators Research, 23(4) Dec. 1990; and Social Indicators Research, 30(2-3) Nov. 1993
Juster, F.T. and Stafford, F.P. 1985. Time, Goods, and Well-Being. Ann Arbor, MI: Survey Research Center, The University of Michigan.
Kaspryzk, D., Duncan, G., Kalton, G., and Singh, M. P. 1989 Panel Surveys, New York: Wiley and Sons; and A. P. Copeland and K. M. White. 1991. Studying Families, Newbury Park, CA: Sage Publications.
Kearny, A.T., Tourangeau, R., Shapiro, G. M., and Ernst, L. R. 1993. "Coverage improvement from experimental residence questions," American Statistical Association Proceedings of the Section on Survey Research Methods: 162-167.
Knoke, D. and Kuklinski, J.H. 1982. Network Analysis, Beverly Hills, CA: Sage Publications, Inc.
Kojetin, B. and Miller, L.A. 1993. "The intrahousehold communications study: Estimating the accuracy of proxy responses at the dyadic level," Paper presented at the annual meeting of the American Association for Public Opinion Research, Pheasant Run, IL.
Kojetin, B.A. and Jerstad, S.J. Feb. 1997. The Quality of Proxy Reports on the Consumer Expenditure Survey. Paper presented at the Society of Consumer Psychology Conference, St. Petersburg, FL.
Kojetin, B.A. and P.A. Mullin. 1995. "The quality of proxy reports on the Current Population Survey (CPS)," Paper presented at the Annual Convention of the American Association for Public Opinion Research, Fort Lauderdale, Florida
Kojetin, B.A., Burnbauer, L.D., and Mullin, P. 1995. Family members knowledge and agreement about each others activities. Paper presented at the annual meeting of the American Psychological Society, New York City, New York
Kormendi, E. 1988. "The quality of income information in telephone and face-to-face surveys." In R. M. Groves, et. al., (Eds.), Telephone Survey Methodology, New York: Wiley, 341-356.
Krosnick, J.A. and Alwin, D.F.. 1987. "An evaluation of a cognitive theory of response-order effects in survey measurement," Public Opinion Quarterly, 51(2), 265-267
Laumann, E.O., Gagnon, J.H., Michael, R.T., and Michaels, S. 1994. The Social Organization of Sexuality: Sexual Practices in the United States. Chicago: University of Chicago Press.
Little, R.J. and Rubin, D.B. 1987. Statistical Analysis With Missing Data, New York: Wiley & Sons
Lord, C.G. 1980. "Schemas and images as memory aids: Two modes of processing social information," Journal of Personality and Social Psychology, 38(2), 257-269.
Madala, G.S. 1983. Limited-Dependent and Qualitative Variables in Econometrics, New York: Cambridge University Press.
Martin, E. and Polivka, A.E. 1995. Diagnostics for Redesigning Questionnaires. Public Opinion Quarterly, 59, 547-567
Martin, E.A., and de la Puente, M. 1993. "Research on Sources of Undercoverage Within Households," American Statistical Association Proceedings of the Section on Survey Research Methods, 1262-1267.
Mathiowetz, N. and Groves., R.M. 1985. "The effects of respondent rules on Health survey reports," American Journal of Public Health, 75, 639-644
McKay, R.B. 1993. "Undercoverage of Hispanics in household surveys," Monthly Labor Review, 166, 38-42.
McKay, R.B. and de la Puente, M. 1995. "Cognitive Research in Designing the CPS Supplement on Race and Ethnicity," Bureau of the Census Proceedings of the Annual Research Conference
McKay,R.B., Stinson, L., de la Puente, M., and Kojetin, B. 1996. "Interpreting The Findings of the Statistical Analysis of the CPS Supplement on Race and Ethnicity," Proceedings of the Bureau of Census' 1996 Annual Research Conference, Rosslyn, VA
McLanahan, S. and Garfinkel, I. 1996. Memorandum to the Working Group on the Methodology of Survey Research on Fathers.
Menon, G. 1994. "Judgments of Behavioral Frequencies: Memory Search and Retrieval Strategies." In N. Schwarz and S. Sudman (Eds.), Autobiographical Memory and the Validity of Retrospective Reports (pp. 161-172), New York: Springer-Verlag
Menon, G., Bickart, B., Sudman, S., and Blair, J. 1995. "How well do you know your partner? Strategies for formulating proxy-reports and their effects on the convergence to self-reports," Journal of Marketing Research, 32, 75-84
Miller, L.A. and Tucker, C. 1993. "The Intrahousehold Communications Study: Family cohesion and the level of knowledge about expenses," American Statistical Association Proceedings of the Section on Survey Research Methods, 942-947
Miller, L.A. and Tucker, C. 1993. "The Intrahoushold Communications Study: A typology of family cohesion," American Statistical Association Proceedings of the Section on Survey Research Methods, 936-941;
Miller, L.A. and T. Downes-Le Guin. 1989. "Improving comprehension and recall on the Consumer Expenditure Interview Survey: Discrepancies in comprehension and recall as a source of nonsampling error," American Statistical Association Proceedings of the Section on Survey Research Methods, 502-507
Mohadjer, L. 1988. "Stratification of Prefix Areas for Sampling Rare Populations," In R. Groves, P. Biemer, L. Lybert, J. Massey, W. Nicholls, and J. Waksbert (Eds.), Telephone Survey Methodology, New York: Wiley & Sons
Moore, J.C. 1988. "Self/proxy response status and survey response quality: A review of the literature," Journal of Official Statistics, 4, 155-172
Moore, J. and Marquis, K. 1989. "Using Administrative Record Data to Evaluate the Quality of Survey Estimates," Survey Methodology, 15, 129-143.
Morgenstern and Barrett. 1974. "The retrospective bias in unemployment reporting by sex, race, and age," Journal of the American Statistical Association, 15, 129-143
Morton-Williams, J. 1993. Interviewer Approaches, Aldershot: Dartmouth Publishing Company Limited
Mosher, W.D., Pratt, W.F, and Duffer, A.P., II. 1994. "CAPI, Event Histories, and Incentives in the NSFG Cycle 5 Pretest." Proceedings of the Section on Survey Research Methods, American Statistical Association, Toronto.
Mullin, P.A., Cashman, E. R., and Straub, H. R. 1996. "Answering question sequences: Attention switching and memory organization." In Hermann, D., McEvoy, C.., Hertzog, C., Hertel, P., and Johnson, M. Basic and Applied Memory: Research: Practical Applications, Volume 2, Hillsdale, NJ: Erlbaum.
Mullin, P.A. and Tonn, B.E. 1993. "Factors associated with proxy knowledge of employment-related information," American Statistical Association Proceedings of the Section on Survey Research Methods
Nargundkar, M.S. and Gower, A.R. 1991. "Cognitive Aspects of Questionnaire Design: Business Surveys Versus Household Surveys," Proceedings of the 1991 Census Annual Research Conference
Parsley, T.L. 1993. "Report on 1990 NHSDA-Census Match: Final Report." Research Triangle Institute: Report to the Substance Abuse and Mental Health Services Administration (Contract No. 271-89-8333).
Peterson, L.S. and Schechter, S. 1995. "Methods to Improve Recall and Reporting in a CAPI Survey: The National Survey of Family Growth." Proceedings of the U.S. Census Bureau Annual Research Conference.
Peterson, L.S. and Schechter, S. 1995. "Methods to Improve Recall and Reporting in a CAPI Survey: The National Survey of Family Growth." Proceedings of the U.S. Census Bureau Annual Research Conference.
Poister, T. H. 1978. Public Programs Analysis, Baltimore, MD: University Park Press
Raghunathan, T.E., Groves, R.M., and Couper, M. 1996. "Statistical adjustment models for unit nonresponse," Paper presented at the 7th International Workshop on Household Survey Non-Response, Rome Italy; and P. Vaisanen. 1996. "Nonresponse in cluster sampling and adjustment of weights in unit level: Application to a household based LFS," Paper presented at the 7th International Workshop on Household Survey Non-Response, Rome, Italy
Riessman, C.K. 1993. Narrative Analysis, Newbury Park, CA: Sage Publications, Inc.
Rogers, T.F. 1976. "Interview by telephone and in person: Quality of responses and field performance," Public Opinion Quarterly, 4040(1), 51-65
Schuman, H. and Presser, S. 1981. Questions and Answers in Attitude Surveys, San Diego: Academic Press, Inc.; and R. M. Groves. 1989. Survey Errors and Survey Costs, New York: John Wiley.
Schuman, H. and Presser, S. 1981. Questions and Answers in Attitude Surveys, San Diego: Academic Press, Inc.
Schuman, H. and Presser, S. 1981. Questions and Answers in Attitude Surveys, San Diego: Academic Press, Inc.
Schwarz, N. and Sudman, S. (Eds.). 1993. Autobiographical Memory and the Validity of Retrospective reports, New York: Springer-Verlag
Schwarz, N. and Sudman, S. (Eds.). 1993. Autobiographical Memory and the Validity of Retrospective Reports, New York: Springer-Verlag
Second Conference on Cognitive Aspects of Survey Methodology, June 1997.
Seltzer, J. and Brandreth, Y. 1994. "What Fathers Say About Involvement With Children After Separation." Journal of Family Issues, 15: 46-77.
Silberstein, A.R. 1989. "Recall effects in the U.S. Consumer Expenditure Interview Survey," Journal of Official Statistics, 5, 125-142
Silberstein, A.R. 1993. "Part-set cueing in diary surveys," American Statistical Association Proceedings of the Section on Survey Research Methods, 398-403
Silberstein, R. 1989. "Recall effects in the U.S. consumer Expenditure Interview Survey," Journal of Official Statistics, 5, 125-142
Sirkin, M.G. and Casady, R.J. 1988. "Sampling variance and nonresponse rates in dual frame, mixed-mode surveys." In R. Groves, P. Biemer, L. Lyberg, J. Massey, W. Nicholls, and J. Waksberg (Eds.), Telephone Survey Methodology, New York: Wiley and Sons.
Smock, P.J. and Manning, W.D. October 1996. "Nonresidential Parents' Economic Ties to Children: New Evidence from the Panel Study of Income Dynamics." Paper presented at the NIH Conference on Father Involvement, Bethesda,
Sonenstein, F.L. and Calhoun, C.A. 1990. "Determinants of Child Support: A Pilot Survey of Absent Parents." Contemporary Policy Issues 8: 75-93.
Sorensen, E. 1996. "A National Profile of Noncustodial Fathers and Their Ability to Pay Child Support." Manuscript, 1996.
Sudman, S. and Bradburn, N. 1973. "Effects of time and memory factors on response in surveys," Journal of the American Statistical Association, 68(344), 805-815
Sudman, S., Bradburn, N.M., and Schwarz, N. 1996. Thinking about Answers: The Application of cognitive Processes to Survey Methodology, San Francisco: Jossey-Bass.
Sudman, S. Bradburn, N.M., and Schwarz, N. 1996. Thinking About Answers: The Application of Cognitive Processes to Survey Methodology, San Francisco: Jossey-Bass.
Sudman, S. and Ferber, R. 1974. "A Comparison of Alternative Procedures for Collecting Consumer Expenditure Data for Frequently Purchased Products." Journal of Marketing Research, 11, 128-135
Sudman, S., Bickart, B., Blair, J., and Menon, G. 1994. "The effect of participation level on reports of behavior and attitudes by proxy reporters." In N. Schwarz and S. Sudman (Eds.) Autobiographical memory and the validity of retrospective reports, New York: Springer-Verlag
Sweet, E.M. 1994. "Roster research results from the living situation survey," Proceedings of the Bureau of the Census Annual Research Conference, 415-433.
Tanur, J.M. (Ed.). 1992. Questions about questions: Inquiries into the Cognitive Bases of Surveys, New York: Russell Sage Foundation
Tanur, M.J. 1992. Questions about questions: Inquiries into the Cognitive Bases of Surveys, New York: Russell Sage Foundation
The National Health and Social Life Survey (Laumann, Gagnon, Michael and Michaels (1994), The National Survey of Men (Tanfer, Billy and Grady, 1993), The National Survey of Adolescent Males (Sonenstein, Pleck and Ku, 1989) and The AIDS Surveys (Catania et al, 1992)
Tucker, C., Casady, R., and Lepowski, J. 1991. "An evaluation of the 1988 Current Point-of-Purchase CATI Feasibility Test." Proceedings of the Section on Survey Research Methods, American Statistical Association, 508-513.
Tucker, C. 1996. "Methodological Issues Surrounding the Application of Cognitive Psychology in Survey Research," Paper presented at the International Conference on Social Measurement, University of Essex.
Tucker, C., Casady, R., and Lepowski, J. 1991. "An evaluation of the 1988 Current Point-Of-Purchase CATI Feasibility Test." Proceedings of the Section on Survey Research Methods, American statistical Association, 508-513
Tucker, C. 1992. "The estimation of instrument effects on data quality in the Consumer Expenditure Diary Survey." Journal of Official Statistics, 8:41-61.
Tucker, C., Casady, R., and Lepowski, J. 1991. "An evaluation of the 1998 Current Point-of-Purchase CATI Feasibility Test." Proceedings of the Section on Survey Research Methods, American Statistical Association, 508-513
Tucker. 1996. "Methodological Issues Surrounding the Application of Cognitive Psychology in Survey Research," Paper presented at the International conference on Social Measurement, University of Essex; G. King, R. O. Keohane, and S. Verba. 1994. "Designing Social Inquiry," Princeton, NJ: Princeton University Press
Turner, C., Forsyth, B., O'Reilly, J., and Miller, H. 1996. "Automated self-interviewing in surveys," Paper presented at the International Conference on computer-Assisted Survey Information Collection, San Antonio, TX.
U.S. Bureau of the Census, Department of Commerce, Statistical Abstract of the United States: 1996, October, 1996.
U.S. Department of Education, National Center for Education Statistics, Pursuing Excellence, NCES 97-198, Washington D.C.: U.S. Government Printing Office, 1996.
Warriner. 1991. "Accuracy of self-reports to the burdensome question: Survey response and nonresponse error trade-offs. Accuracy of self-reports, " Quality and Quantity, 23:253-269.
Willis, G.S., Royston, P., and Bercini, D. 1991. "The Use of Verbal Report Methods in the Development and Testing of Survey Questionnaires," Applied Cognitive Psychology, 5, 251-267
Yin, R.K. 1989. Case Study Research: Design and Methods, Newbury Park, CA: Sage Publications, Inc.
1. Second Conference on Cognitive Aspects of Survey Methodology, June 1997.