Trends in Voluntary Group Membership: Comments on Baumgartner and Walker
Total Page:16
File Type:pdf, Size:1020Kb
Trends in Voluntary Group Membership: Comments on Baumgartner and Walker Tom w. Smith NORC University of Chicago GSS Methodological Report No. 60 February, 1989 This research was done for the General Social survey project directed by James A. Davis and Tom W. Smith. The project is funded by the National Science Foundation Grant SES-8745227. I would like to thank Jane Junn for data runs from the 1967 Verba Nie study, Mary E. Morris and P. Luevano for questionnaires from the 1952 and 1985 American National Election studies (ANES) , Santa Traugott for an ANES memo, and the Roper Center for the 1952 AIPO study. I would also like to thank Jane Junn, David Knoke, James A. Davis, Richard Niemi, Christopher Walsh, and Roger Tourangeau for comments on an earlier draft on this paper. Baumgartner and Walker ( 1988) argue that participation in voluntary associations has increased since the 1950s and that flaws in what they refer to as the Standard Question on group membership distort the time series and prevent the true expansion of group membership from being detected. This note examines the evidence on trends in voluntary group memberships and evaluates their critique of the standard Question. Baumgartner and Walker present three pieces of evidence in support of the notion that voluntary group membership has increased: 1) many new groups have formed and grown rapidly during recent decades, 2) national surveys of group membership show a rise in memberships from 1952 to 1974, and 3) inadequacies in the Standard Question mask further rises in membership since 1974. Case Studies of Group Membership First, they argue that monographic studies of membership growth indicate "the rise and development during the past half century of movements promoting new political causes (p. 909)." These studies include "numerous stories of explosive growth not only in the numbers of organizations, but also in the number of people participating in group activities (p. 909) . 11 They conclude that "the published studies of group development constitute strong circumstantial evidence of growth in the membership of interest groups ·in the years since World War II (p. 909). 11 The problem with this analysis is that both individual memberships in groups and groups themselves have high turnover (Babchuk and Booth, 1969; Hannan and Freeman, 1988). The formation and growth of certain groups is always to a greater or lesser extent off-set by the decline and demise of other groups. Baumgartner and Walker acknowledge this fact ( 11 it is possible that there were declines in membership in other areas that went unreported (p. 909; see alsop. 912)"), but dismiss this factor to reach their conclusion of notable increases in group membership. However, many declines in group memberships have actually been widely reported. To cite only two counter examples, membership in the International Ladies Garment Workers Union declined from 363,000 in 1971 to 173,000 in 1987 and the United Methodist Church adherents dropped from 11,100,000 in 1965 to 9,400, 000 in 1982. These trends have been widely reported in numerous discussions of the decline of the American labor movement (Cornfield, 1986; Lipset, J,986; Gifford, 1984) and of mainline Protestant churches (Roof and McKinney, 1987; Smith, 1988b; Hoge and Roozen, 1979) . Such notable declines seriously question the presumption that group memberships in general have been rising. Of course the decline of these particular groups tells us no more about changes in overall voluntary association membership than the rise of the groups cited by Baumgartner and Walker. Ultimately if we want to know whether group membership in general has been increasing, we have to study group membership in 1 general. National Surveys of Group Membership Next, Baumgartner and Walker do exactly that, they turn to national surveys of voluntary group membership. In their Figures 1 and 2 they examine respectively the percent reporting one or more group memberships and the mean number of memberships in "comparable national surveys (pp. 910, 911)." The basic problem with their analysis is that these surveys donot actually contain comparable measures of group membership. As the Appendix indicates, until the 1974 General Social Survey (GSS) no two observations are the same and the various questions differ in numerous notable ways .1 Table 1 indicates some of the major differences between these questions. Perhaps the most important difference is how one inquires about group memberships. Both the survey research and cognitive psychology literatures on recognition and recall (Crowder, 1976; Tourangeau, 1984) indicate that the number of events reported will vary directly with the specificity and directness of the question. The number reported will be lowest when only a single general inquiry is used with no examples and no prompts; medium when examples are verbally mentioned and/or listed on a show card; and highest when each type of event is inquired about separately. 2 The research on group memberships indicates that this general pattern applies specifically in the case of membership in voluntary associations (Hyman and Wright, 1971; Taylor, 1975; · 1976; Smith, 1988b; Nelson, Baker, and Nelson, 1977; Babchuk and ·Booth, 1969; Knoke and Thomson, 1977; Thomson 1 The appendix and Table 1 include both the questions used by Baumgartner and Walker and similar national survey questions. 2 It also appears that a question that solely focuses on one event/feature will record more mentions than a question that asks about the event as part of series of related events even when the event is asked about separately in both cases. For example, the GSS and the 1972 ANES ask about union membership both as part of their general membership question anci. as a separate question.- Union membership is 5-6% higher on the stand-alone question than on the general question. The literature on labor union membership is also very instructive about alternative definitions of memberships and the differences that occur when different measurement procedures and definitions are used (U.S. Bureau of the Census, 1975; Adams,l985; Gifford, 1984). 2 and Knoke, 1980) .3 Other notable differences in question wordings occur in 1) the area referred to (community vs. all groups), 2) the use of probes to encourage more mentions, 3) references to personal vs. family memberships, 4) miscellaneous other factors such as restrictive definitions of groups and limited space for recording the name of groups. One example of the problem is the differences between the NORC studies in 1953 and 1958. These studies were first analyzed by Hyman and Wright (1971; Wright and Hyman, 1959) and it is thus not surprising that Baumgartner and Walker would include them as comparable. In fact, however, the two questions differ notably. The 1953 question refers to "anyone in the family," while in 1958 the reference is to "adults in the family." Also, the 1953 question consists on a single, general group question, while the 1958 item combines together a specific question on unions with a follow-up question on "other organizations." Finally, on the 1958 study 8% of the respondents were excluded from analysis apparently because they didnot answer the follow-up question, while few missing cases occurred in 1953. Hyman and Wright argue (1971, p. 196) that these non-respondents probably did not differ in their group membership level from the respondents, but offer no evidence in support of this assumption. Give:t:1 these, as well. as other differences, any direct comparison of membership levels ·seems j,mprudent. · A second example is the comparison between the I967.Verba Nie study of political participation and. the 1974-1988 GSS questions. Baumgartner and Walker compare these questions witnout making important adjustments for wording and possible context effects. While the GSS question was modelled after·the Verba=Nie item and these questions are probably more . similar than any others in BaumgartnerjWalker time series, they differ iri several notable ways. First, as previous researchers have noted, the GSS records significantly more members of Church-Affiliated Groups, because the GSS added this group to its set of specific inquiries. For overtime comparisons the standard adjustment for this difference has been to exclude church groups from both 3 One might try a personal experiment to illustrate this point. Ask yourself, "What voluntary groups/organizations do I belong to?," and give yourself two minutes to answer (more time than the average survey respondent would have to answer) . Then go through any of the longer lists such as on the 1972 ANES or GSS questions. You will probably come up with more memberships. Finally, think about the issue of memberships for a day or two, check your CV, review what newsletters and magazines you receive, and talk to your spouse (if appropriate). You should come up with even more memberships. Instantaneous and unaided recall of group memberships is a difficult and fallible task. 3 surveys (Knoke and Thomson, 1977; Nelson, Baker, and nelsom, 1978; Nelson, Nelson, and Baker, 1979; Thomson and Knoke, 1980; Nie, 1988). When either this standard adjustment is made or an alternative adjustment which also takes into consideration differences in the indicators of political groups, we see that the apparent increase in group memberships between 1967 and 1974- 1988 disappears (Table 2) .4 Besides these differences in question wordings there is the real possibility that context effects may also distort the BaumgartnerjWalker time series. Early investigators of voluntary membership have noted possible variation in the level of group membership due to context effects (Hausknecht, 1962; Hyman and Wright, 1971) . While context effects can come in many forms (Schuman and Presser, 1981; Smith, 1986; Tourangeau, 1984), the most likely in the case of group membership would be a priming effect by which prior questions related to group membership would facilitate recall and increase the number of groups mentioned.