THE URBAN COMMUNES DATA SET: SAMPLING FRAME AND DATA COLLECTION DETAILS
Adapted from Alienation and Charisma, by Benjamin Zablocki.
Design Aspects
The UCDS itself is a multiwave, multimethod study. While questionnaire techniques predominated, close to 1,500 hours of person‑to‑person, open‑ended interviews were recorded on tape and on interview schedules, and these formed an indispensable part of the data library. Although opportunities for prolonged participant observation were few, these were utilized whenever possible.
The following is a brief description of the basic survey instruments used. Titles that are in bold indicate that most data from this instrument from the first wave is being made public.
I. The Commune‑Level Data Protocol: This is a schedule of items that uses the commune as the unit of analysis. The protocol includes fifty‑four closed‑ended items (such as dates of founding, type of authority, rationality of decision making, economic organization, method of child rearing, type and intensity of ideology, and marital and sexual arrangements) and was filled out by fieldworkers in cooperation with members of the communes themselves.
II. Personal Background and Participation (Long Form). The "long form" is a schedule of items that explores in detail the individual's personal biography. The long forms were personally administered by fieldworkers to four or five members of each commune. These interviews took from two to four hours to complete, and they cover—rather extensively—each respondent’s early life, with particular emphasis on the year just prior to joining the commune. The long form also devotes considerable attention to the respondent's current life, present goals, and participation in the ongoing activities of the commune. This schedule is a combination of open‑ and closed‑ended items.
III. Personal Background and Participation (Short Form). It was recognized, in advance, that not all commune members would have the time or the interest to respond to the detailed questions of the long form. Consequently, a "short form" instrument was developed that contains twenty of the most important closed‑ended questions from the long form. The short form comprises the minimal data set desired for each commune member.
IV. Supplementary Questionnaire. Preliminary analysis of data from the “five city” 1974 summer survey indicated that several important areas of communal life (e.g. decision making, personal goals and expectations, communal change) invited deeper exploration than that provided in the original five instruments. As a result, a supplementary questionnaire was developed to replace the "short form" and was utilized in the out‑of‑phase Boston data collection during the winter of 1975. The supplementary questionnaire includes all the items from the short form (so that short‑form material is uniform and comparable across all three personal background instruments).
V. Attitude Questionnaire. The attitudes instrument is a set of ninety‑nine Likert-type items drawn, for the most part, from standard scales used on national surveys for the assessment of alienation, preference, self‑esteem and self perception. Priority for inclusion in the questionnaire was given to sets of items such as the Srole anomia and Harris Alienation scales, both of which have been widely administered to other, noncommunitarian samples of the national population. Only a subset of these are being made public though the remaining items are available upon request.
VI. Relationship Questionnaire. This instrument consists of sociometric choice questions in which each individual is asked to respond to a complete set of questions describing his/her relationships with every other member of the commune.
VII. Ex‑Members. Because the very nature of communitarianism tends to inhibit the development of firm, long‑term personal commitments, a high turnover rate was anticipated between wave one and wave two. Follow‑up of ex‑members, therefore, comprised an important segment of the research. During the course of 1975 summer data collection, fieldworkers compiled lists of ex‑members, and, in cases where the ex‑members could be located within the SMSA, fieldworkers distributed special questionnaire forms to these individuals. In most cases, however, ex‑member location procedures took several steps, often involving contact with a member's parents or other relatives living in different states (these contact addresses were obtained from all respondents during the 1974 data collection process). Because of budget and time constraints, it quickly became apparent that not all ex‑members could be traced and interviewed. As a result the ex‑members were divided into three priority categories, with major search and retrieval efforts being devoted to those with high priority and minimal search and retrieval efforts devoted to those with low priority. In all cases, however, where a respondent could be readily located, no matter what the priority rating, efforts were made to retrieve ex‑member information. Completed questionnaires were eventually received from 32 percent of the ex‑members. Systematic biases were observed when comparing ex‑member respondents and nonrespondents on the basis of their first‑wave responses. Those who did not return ex‑member questionnaires were quite a bit more radical in their first wave responses than those that did.
Sampling
The 60 communes in the UCDS do not constitute a probability sample of American communes. There was no feasible way to enumerate the population of American communes from which the sample was drawn. Even the total number of communes in this population can only really be estimated to an order of magnitude. Nevertheless, considerable efforts were made to eliminate the known sources of bias that may have skewed the results of previous commune studies. It is probably safe to say that inferences drawn from the results of this study, to American communes in general, will usually be closer to the mark than inferences drawn from any other currently published study. However, the reader interested in drawing such inferences would do well to take heed of the sampling idiosyncracies described in the following paragraphs. We begin with the definition of “commune” that guided the sampling, then describe how regions were selected, how cities were selected within regions, and finally how communes were selected within cities.
The Population
A commune was defined as follows: it is a household in which
1) Five of more adult individuals live (plus children if any);
2) If no children are present, both sexes must be represented among the adults (to eliminate monasteries and related organizations);
3) The majority of dyadic ties between these adults are not kin or marriage ties;
4) Their joint residence is a result of their choice, without compulsion, for an indefinite period of time;
5) Their reason for establishing the household is primarily to reach some ideological goal having to do with the achievement of community;
6) The group must have a collective identity known to outsiders (e.g. name, common function). While this condition was introduced largely for methodological reasons (secret communes would not be learned about and hence would not enter the sample), it also ensured that all groups were more than opportunistic collections of roommates;
7) The commune was in existence in July 1974.
In most instances, this definition was sufficient to allow us to determine unambiguously whether or not a given observable entity was eligible to be included in the sampling frame. However, there were a few cases in which the decision to include or not to include was based on judgments not explicit within our definition. In cases in which data collection began on a group that was later disqualified from analysis, we have retained that data for secondary analysts. Thus we have extensive network data on one group later determined to be basically involuntary (a rehabilitational commune in which entrance was a court-prescribed alternative to jail time).
A decision was made to include all communes in existence during the time of the enumeration even if they were begun in an earlier era. This complicated our attempts to treat the commune sample as part of a single social movement. However, it served to increase the variance in commune duration, which proved to be useful. Because no effort was made to select communes on the basis of founding‑year cohort, this decision could have gone either way.
A decision was made to exclude groups of people who had made definite plans to live communally (even if these included formal contractual agreements) but who were not yet doing so. Groups that had occupied the same piece of land as another commune but had revolted from that commune to set up their own were also excluded unless the schism was recognized as permanent by both groups.
A decision was made to treat each household of a multihousehold communal federation as a separate commune. However, in some instances even the definition of what constituted a single household proved ambiguous. Household was defined as a functional rather than as a physical entity. A group owning or renting several separate dwelling units and freely and fluidly distributing its members among them was considered to be a single commune. If interhouse access was not free and fluid but limited in the service of ideological norms, the commune was still treated as a single unit. Examples of the latter were communes with a separate house for the charismatic leader, as a distancing mechanism, or separate houses for men and women or for the fully initiated and the novices. It followed from these decisions that the inclusion of one commune from a communitarian federation did not reduce the probability that another commune from the same federation in another geographical region would also be chosen. This happened in several instances. In one of them, as shall be discussed shortly, it was deliberately contrived to happen.
In a second instance, however, groups originally considered to be one household were determined to be independent were later divided. This leads to the apparent “birth” of a commune after data collection had begun. An indicator variable (“HEADLAND”) identifies all the households in this group.
Selection of Geographical Regions
The commune sample is made up of ten communes from each of twelve distinct geographical areas. Each of the twelve areas is a cluster of contiguous counties. Six of the county clusters are rural and were discovered and delineated through field work. The other six county clusters are urban and follow metropolitan boundaries established by the U.S. Census Bureau.
Urban communes were to be found in virtually all large‑ and medium‑sized cities and their surrounding suburban areas. Rather than aiming for maximum density of concentration, we aimed to maximize geographical diversity. The U.S. Census Bureau divides the nation into nine major geographical regions. Originally, two of the regions were combined (South Atlantic and East South Central) and the largest Standard Metropolitan Statistical Area from each of the eight regions was selected for study. However, pretests clearly indicated that time and budget constraints would not permit detailed ethnographic and survey analysis of all eight areas. As a result, Chicago and Denver (East North Central and Mountain regions) were arbitrarily dropped from the list and the following six SMSAs remained: Los Angeles, Twin Cities, Atlanta, Houston, New York, and Boston.
The primary reason for selecting cities by arbitrary census definitions was to avoid the regional bias that has affected previous studies of communal life. Because most of the active commune researchers in recent years have been based on the West Coast, a disproportionate amount of research on contemporary communes has focused on communities west of the Mississippi. Although there are probably more communes in the West than in the East, this regional bias in sampling has deprived researchers of important systematic data concerning differences in communal life in different regions of the country.
The Enumeration Stage in the Urban Areas
During a preliminary month of fieldwork in the summer of 1974, fieldworkers in each city compiled a comprehensive[*] census of communes within the SMSA. This compilation was achieved through the exploitation of every foreseeable (and serendipitous) source of commune information. Sensitive to the sampling biases that have plagued most commune research (e.g., undercounting of "nameless" communes, overrepresentation of highly institutionalized groups, underrepresentation of short‑lived and nonmiddle‑class groups), fieldworkers were instructed to start their census work from a wide variety of different entry points.
The search for communes was, of course, complicated by definitional problems. Many groups that satisfied the study's definition of "commune" did not, in fact, answer to the name of "commune," and, conversely, many groups that called themselves "communes" did not satisfy the definition at all. In explaining their needs to informants, therefore, fieldworkers used the word "commune" with discretion and emphasized instead the ideological, the common residence, and the multiple‑person characteristics of the groups that they were seeking.
Sampling of Communes
After approximately four weeks of field work, the census information from each of the cities was sufficiently complete to permit selection of individual communes to be studied in depth. Although a random sampling of the enumerated universe would have provided the most stringent sampling criterion, much of the representative control of the six‑city sample would have been lost by random selection. Instead, the individual communes were selected on the basis of certain key variables such as ideology type, population size, number of children, type of neighborhood, and year founded. In the rare instances where access was flatly denied by a selected commune, the next highest group on the priority list was chosen for study.
There was, however, one significant deviation from the "relevant variables" selection procedure. Early in the fieldwork, a number of nationwide religious cults, organized in the form of federations of communal households (whose members often moved among households in different states), were located. In fact, ten of these cults were identified during the summer field work of 1974, each centering around the veneration of a guru, prophet, or avatar. Because of the obvious sociological and historical importance of these new religious federations, and because of the unique opportunities they provide for cross‑regional comparisons among representatives of the same communitarian organization, one cult was selected for study. Representative households belonging to this cult were included in the sample from each of the six cities. A group-level variable (“GURULAND”) indicates whether a commune was part of this federation or not.
Unfortunately, it is difficult to assess the representativeness of this set of communes, for there is virtually no information available (other than this study's six‑city census) concerning the number or types of communes existing in the United States. The very few systematic studies that have been conducted on communes have been limited to small geographic areas (e.g., New Haven and Boston) and have explicitly excluded from consideration "creedal households" and federations. By excluding religious, spiritual, and political groups (and by concentrating, thereby, solely on "domestic households"), these earlier studies contribute little to an overall estimate of the commune universe.
In summary, the commune sample deviates in many ways from a strict probability sample and requires far more cautious interpretation of specific results. However, it allows us some latitude in making inferences about the population of communes and commune members as a whole, while giving us a greater flexibility than a probability sample would provide.
Selection of Respondents
An attempt was made to study the entire adult (over 15 years of age) population of each of the communes. However, not all were willing to participate, and not all of those who did participate took part in all aspects of the data collection.
The six survey instruments already discussed were used as indicated. In addition to strictly individual‑level data, a complete enumeration of dyadic relationships on a wide variety of relational variables was elicited. Systematic global data concerning leadership, economic organization, marriage, child rearing, decision making, and so on were also collected for all rural and urban communes. Ecological data on each of the communes were collected for the surrounding neighborhood, i.e. the census tract.
Data Collection
Because of the wide variety of emotional and rational justifications for denying access to strangers with tape recorders and survey instruments, fieldworkers spent a great deal of time with individual members and with the groups as a whole before they sprang the question of official access. A wide range of objections were voiced by individual commune members:
"This is our home, our family, not some scientific laboratory experiment"; "How can you possibly understand anything about what we're up to with a bunch of printed questions"; "We've been through this before. Some guys came and lived here for almost a month last year and got in everyone's way and asked a lot of questions and said they were writing a book and we have never seen one word of anything they wrote and have never heard from them again"; "A lot of people have been busted around this city lately and I think we'd better keep our mouths shut"; or, "I have a friend who lives in a commune who just lost custody of her child because of it, and I'm not about to put myself in the same position."
But, despite these objections and despite the self‑imposed research constraint that the commune as a whole must give permission to be studied (although individual members might refuse), only six of the originally selected communes denied access. Furthermore, during the course of the first‑wave research, only one of the groups that had agreed to participate withdrew that agreement (and that one only prohibited questionnaire distribution) once data collection was underway. Others, however, forbade the use of certain specific questionnaires. The attitude questionnaire was most frequently griped about but it was the relationship questionnaire that was most frequently forbidden or restricted. Five communes refused to allow the first wave relationship questionnaire to be distributed on principle.
The extent of cooperation and the apparent quality of the data collected were, on the whole, quite high. All the communes participated in the gathering of the systematic commune‑level information. At the individual level there was a more selective but still quite high rate of response. For the first wave, we received personal background data (short or long forms) from 81 percent of the possible urban respondents (N = 667), more than half of whom participated in autobiographical interviews. Relationship data were obtained from 70 percent (80% if we exclude the groups which refused on principle to allow for the distribution of these questionnaires) and attitude data from 60 percent. The lower rate of response to the attitude questionnaire also unfortunately corresponds to a lower quality of data. Ethnographic and interview cross‑checks revealed a uniformly high quality to the responses to personal background and relational instruments. These were also generally judged to be relevant by the commune members. The attitude questionnaire met with much greater hostility and probably produced a somewhat higher rate of nonserious response.
From the experiences of the first wave of urban data collection and analysis, it became apparent that certain changes in procedures and instruments would enhance the usability of the second‑wave data (always, of course, with an eye toward comparability between first and second‑wave materials). Perhaps the most significant difference between first‑ and second‑wave urban data collection procedures was a difference in emphasis on units of observation. For the second wave, the communal group as a unit of observation and analysis assumed precedence over the individual‑level biographical concerns of the first wave.
In two of the cities (Minneapolis and Los Angeles) the 1974 resident fieldworkers, who were, by this time, well known to most of the groups, also conducted the 1975 fieldwork. In Minneapolis, none of the ten groups refused reentry permission. In Los Angeles, two groups refused reentry. One of the refusing groups had also denied access in 1974 for all but tape‑recorded interviews. The other group had participated fully in 1974 but did not grant access in 1975.
Problems of Measurement
Reliability and Validity
Two kinds of reliability were matters of concern: response stability and intercoder reliability. With respect to the first, attitude and relationship responses were particularly open to question. The test‑retest reliability of the attitude questionnaire was measured on a group of college students with a 48 hour interval between tests. On no item did more than 10 percent of the students shift from agreement to disagreement or vice versa, and for a majority of items there were no such shifts. However, higher rates of shift, up to 25 percent, were recorded when a shift was defined as any change along the five‑point Likert scale. Shifts into and out of the no answer/no opinion category were particularly frequent. The attitude items should be considered, at best, of marginal reliability. The stability of the relational responses was harder to check, requiring as it did a test group of people who knew each other quite well. The only test‑retest reliability check done on this instrument used the commune project office staff, tested in the morning and again in the afternoon. Almost no shifts were observed between tests, indicating a reasonable level of stability. However, more work on this instrument needs to be done.
Intercoder reliability was more of a problem, particularly for the global‑level urban commune data recorded in the field. For such judgment items as ideological intensity and rationality in decision making, for example, it was difficult to get a farflung staff of fieldworkers to apply common yardsticks. In the end, after many hours spent in long‑distance telephone conferences, only a series of around ‑the ‑country tours by the core staff (Zablocki, Bradley, and Aidala) to visit, at least briefly, virtually all the urban communes, assured a degree of data comparability that we felt we could live with. There is still, however, undoubtedly more error of this kind left in the urban data than in the rural, where at least a single person's judgment prevailed.
Evidence for measurement validity was provided by the multimethod approach to data collection as we discussed earlier. In general, the survey measures of key concepts converged quite well with nonsurvey measures. Specific validation problems are best discussed in terms of the specific concepts for which the problems emerged. The following are some of the more important examples:
1. Relational measures: At the dyadic level, relationship questionnaire items converged with field observations and the opinions of interviewees. But, to the extent that these responses were used to trace a network of an entire commune, the data were quite vulnerable to incompleteness of response. Further, unlike the other questionnaires, the probability of filling out a relationship questionnaire was judged to be far from independent of one's position in the network.
2. Power: The evident validity of our naive measure of interpersonal power was quite surprising. The indicator was based upon a single question asking each respondent to evaluate his or her relative power with respect to each other commune member. A strict power relationship was said to exist between A and B if A claimed power over B and B acknowledged A's power over B. A relaxed power relationship was said to exist if A claimed power over B and B did not claim power over A, or if B acknowledged A's power over B and A did not deny having power over B. Both measures produced revealing network diagrams, with almost no intransitive cycles.
3. Dyadic partiality: Four measures were used, at one point or another in the study, to ascertain the extent to which commune members singled out one another for a special relationship. It was, of course, important to keep this measure free of overlap with measures of emotional attraction or repulsion or with measures of deference or respect‑ The research design originally called for using the amount of time that any two people spent together, just by themselves, as an indicator of dyadic partiality. But this was found to be a variable that was highly constrained by factors having nothing to do with partiality. A useful substitute measure was found in the question, "If the commune did not exist, would you want to have a close relationship with this person?'' However, unfortunately, that question was not asked of many of the early respondents in the first wave.
Further useful indicators were significance and intimate knowledge. The first was determined by asking each person to list the five most significant people in his or her life (in or out of the commune). This was the only question asked in which the respondents were forced to be selective by being limited in the number of names they could list. The second was determined by asking each person if he or she were aware of each other person's father's occupation. Informal discussions with a number of social psychologists indicated that this was a good way of distinguishing people who had spent some time talking with each other about themselves from those who had not. A dyadic relationship of partiality was said to exist if and only if each member of a dyad listed the other as one of his or her five most significant others and if each member also indicated awareness of the other's father's occupation. Density of partiality ranged among the communes from 0 percent to 33 percent.
4. Disintegration: Defining communal disintegration involves three distinct issues: loss of domicile loss of members, and loss of corporate identity. Each has something to do with disintegration, and each may vary independently of the others. This problem has plagued commune archivists for over a hundred years, leading to substantial discrepancies in historical statistics. For the purposes of this study, loss of members is clearly the most important criterion. However, even a complete loss of membership was not by itself deemed sufficient to define a commune disintegration. We define a disintegration to have taken place whenever at least two out of the following three events take place in a year:
(1) 100 percent membership turnover,
(2) change in or loss of domicile;
(3) abandonment of corporate identity.
Although this definition is somewhat arbitrary, it did sort out the problem cases nicely, sorting them into the categories that seemed intuitively right.
