From: MJMARGOL@UCBEH.SAN.UC.EDU Date: Sat, 17 Sep 1994 13:04:30 -0500 (EST) Subject: Survey Report Dear Net User: Thank you for responding to our survey. Below please find the survey report you requested. This report was presented at the Annual Meeting of the American Political Science Association in New York City, September 1-4, 1994. Bonnie Fisher, Michael Margolis, David Resnick. apsa paper corrected version. -1- A New Way of Talking Politics: Democracy on the Internet Ah, the Internet...I remember when the name didn't even exist. Back then, it wasn't politics...it was people exchanging some ideas and programs and such over the net, talking with friends in Israel, or playing an interactive war game, etc. ... POLITICS! of the network...yee gads... Get your society off of our net!!! The Internet has been awfully nice over the years, and now it is being ruined by people who feel it has to be somehow defined, measured, analyzed, have laws made on it, charges assessed against each byte, and on and on and on...WILL YOU PLEASE STOP??? Leave the net as it is. Yeah the Internet has its problems, but the very large majority of us like it *just* the way it IS. I'm sure you're all well intentioned etc....But please, just sit back and enjoy and stop trying to put the labels and constraints of the society we are forced to live in real life onto the Internet. ....Trying, knowingly in vain to stop the world from ruining the net... --An informative refusal to our survey. Introduction: To spam is human. To forgive is divine. To divine a spam operationally, however, presents some difficulties. As does studying democracy on the Internet. As does bridging the gap between the idealistic versions of electronic democracy and the realities of civic life and political participation via cyberspace. But we are getting ahead of our story. The original purposes of our paper were fourfold: first, to describe some models of electronic democracy that rely on the power of the Internet to enhance citizens' participation in politics; second to describe how we developed and applied an instrument to collect data on how subscribers use the Internet to participate in civic life; third, to analyze and report the results of our data collection; and fourth to evaluate the implications of our results for the validity of the models with which we began and to explore some of their consequences for democratic theory . As it happened, our methods had some unintended consequences. Some of these proved valuable for determining the character of civic life on the Internet and its implications for democratic theory. Others had deleterious effects regarding the representativeness of those who responded to our survey. In the sections that follow we consider each of the above purposes in turn. i Democracy and the Internet: Millions of citizens now connect with the Internet, and a significant number of them engage in political discussions (Rheingold, 1993: 8-12). This citizen interaction in cyberspace deserves attention from political scientists, for it has the potential to affect both the formation of public opinion and the conduct of democratic politics. The Net provides new ways for citizens to connect with each other. It has fostered what have been called "virtual communities," groups whose members meet only in cyberspace; yet some of these communities carry on a lively political life. The Internet provides a new public space--an electronic agora if you will-- that facilitates democratic participation in politics adapted to advanced post-industrial societies. Having evolved from a RAND design for a communication, command, and control network that could survive a nuclear attack, the Internet has no center. (Rheingold, 1993: 7-8). Communication in cyberspace differs from broadcasting, therefore, in that there is no central cluster of studios from which most information on the network gets distributed. Each citizen is both a receiving node and a broadcaster. In theory at least, political agendas can emerge through interaction among participants who share equal powers of communication. And, as the Internet extends across national boundaries, the virtual communities provide unique opportunities for citizens to participate in the emerging politics of the global village. Here is a world-wide network, accessible from a computer terminal; the terminal needs only an ordinary telephone line and a modem to provide an opportunity to participate. For optimists, political participation in cyberspace approximates an ideal type of communitarian democracy that emphasizes mutuality. Not only do the citizens share equal powers for receiving and broadcasting, but they also share equal access to vast stores of data. The time and money needed to become informed about any topic drop substantially when citizens can employ "gopher" or "archie" servers to locate and retrieve desired information on a vast variety of topics, including matters of public policy that comprise the formal business of government. Mass democracy, whether conceived of as an adversarial contest among competing interests, a unitary process for building consensus, or some combination of the two, becomes feasible. Civic life of course extends beyond formal matters of public policy. People can interact with one another over a variety of matters, and such interactions can build a sense of community among those who discover shared interests. This sense of community may lead to the formation of virtual communities that operate Usenet bulletin boards, Listserve mailing lists, or even separate community networks, such as the Cleveland Free-Net or TriState Online in Cincinnati. In some cases virtual communities may choose to form distinct conferential networks, such as the WELL. Some of these communities may operate as cooperative societies. That is, members may participate for one another's mutual benefit without the expectation of a quid pro quo for each particular contribution or service they provide. The civic life of the virtual community may resemble the mutuality of a barn raising or a pot luck supper. (See Barber, 1984: 229-33; Mansbridge, 1980: 8-10; Rheingold, 1993: 12-13.) The civic life of cyberspace may have another advantage over the ordinary interactions of civil society. Because people usually interact with one another only by exchanging texts or documents, the common prejudices they hold against particular sexes, races, ethnic or religious groups, etc. are more likely to remain irrelevant. In short, the civic life of the citizens of cyberspace may consist of purer, less bigoted interactions than those that commonly take place among citizens interacting face to face. In this regard, the civic life of cyberspace may represent a higher order of democracy than has been achieved elsewhere. Those who consider notions of virtual communities based on mutuality too idealistic may still perceive cyberspace as an environment that facilitates a more traditional civic life that involves organization, mobilization and bargaining among interest groups. Bulletin boards and mailing lists represent two powerful means of communication among group members. They can be used to increase the efficiency and lessen the overall costs of traditional participation in adversarial politics. Positions can be developed; strategies devised; bargains and compromises achieved. Where traditional democratic politics involve the resolution of group conflict through combinations of pressures, bargaining, and compromise, some like-minded citizens of cyberspace may develop a third type of civic life that involves little or no exchange among those who hold different opinions. The Internet allows virtual communities to develop that resemble the semi-private spaces of modern health clubs more than the public spaces of agoras. Instead of meeting to discuss and debate issues of common concern to the society, members of these virtual communities meet largely to promote their own interests (whether or not these are political) and to reinforce their own like-mindedness. They tend to exclude anyone who disagrees. As a consequence, however, they also reinforce the fragmentation and factionalism of modern society. The public space of Athens certainly facilitated democracy, but we must also recall that Athenian citizenship was a social status, not a universal right. Democracy implies equality, and so it did for the Athenians, but classical political equality existed within a legally and socially stratified society. Citizenship was limited primarily to free male sons of citizens; women, slaves and resident aliens were excluded. The modern analogy suggests a fourth type of technological civic life that requires economic resources and technical skills to participate. While no one argues for recreating the stratified citizenship of Athens on the Internet, the virtual communities may nonetheless reflect biases that favor the richer and better educated. Even though the federal government subsidizes most Americans who spend time cruising the Internet, access is still limited mostly to those connected with educational institutions, research organizations, state and local governmental units, and businesses which also pick up a portion of the tab. These participants tend to be better off and better educated than the average citizens, who, we might add, end up subsidizing their participation, yet another form of American public "wealthfare." Economics, however, is not the only barrier to participation. A significant sector of the population lacks the technical skills to participate. Today's Internet is hardly user- friendly, and we suspect that a great many who participate in its civic life like it that way. Indeed, our experience suggests that if the most technologically sophisticated had their wish, they would draw up strict immigration laws that would limit newcomers ("newbies") to those who would be useful to the Net, as they understand it. A rough equality among the technologically competent seems acceptable, but extending the technology to a hitherto unheard from mass public threatens elite control. Notwithstanding its theoretical potential for increasing political participation, the technology itself may actually be arranged to discourage mass participation. Finally, we must consider a fifth type of civic life that turns the first on its head. Instead of using the relative facelessness of the Internet to overcome prejudice, these participants would use it to disguise their true identity. Instead of propagating truth and fostering mutuality these participants would manipulate opinion, exercise domination, and enhance their own power. They would place ordinary participants under surveillance, censor their communications, restrict their public space. Where advertisers and public relations experts attempt to modify reality by developing positive images of their clients' interests, these participants push the process of modification a step further. They aim to create a virtual reality that other participants cannot distinguish from the civic life of either cyberspace or the real world. Each of the five general types of civic life has some characteristics that we can attempt to measure through survey research. For instance, we would expect those whose participation fits the first general type, mutuality, to report that they use the Internet mostly to establish friendships and social contacts, express opinions, obtain others' opinions, and to organize or plan group activities. We would expect those who participate for more traditional political reasons involving bargaining and compromise among groups to place more emphasis on group organization and exchange of opinions and to place less emphasis on building friendships and social contacts. In contrast, we would expect both those who participate for contact with like-minded individuals or for economic or technological advantage to place more emphasis on establishing professional contacts or interchanges, uploading and downloading information, and possibly seeking entertainment or amusement. We would expect the participants in like-minded groups to cite friendships and social contacts more frequently than would the economic and technological elite. Both types of participants would be less tolerant of social scientists asking them about their usage of the Internet than those whose participation resembled the civic life of the first two types. Both types would tend to subscribe to commercial online services more than would the first two. In comparison to all other types, however, the technological elite should have the longest experience in using the Internet. Survey research would seem less suited to measure the fifth type of civic life, manipulation of virtual reality by an anonymous elite. We present evidence in the next section, however, that suggests that self-appointed censors do indeed exist, though in our case at least, they did not attempt to hide their work so thoroughly as to leave no tracks. The five types of civic life we have outlined: 1) mutuality; 2) democratic bargaining; 3) like-minded exchanges; 4) technological; and 5) manipulation and domination, though analytically distinct, are not mutually exclusive. It is certainly possible for a person to participate in more than one type of civic life. In addition, regardless of their styles of civic life, we expect that those who use the Internet have disproportionately higher incomes and education than the average citizen, that they tend to be younger and more technologically sophisticated than their compatriots, and that their socio-economic characteristics reflect those of the dominant economic and technological elite of western industrialized countries, namely college-educated white males. While we had above typologies crudely in mind when we began, our principal purpose, as we explained to potential respondents, was simply to conduct "an exploratory analysis: a first attempt to provide a general description of how citizens are using computer networks to participate in civic life." We indicated that we were preparing a paper for delivery at the Annual Meeting of the American Political Science Association, and we promised to send results to respondents who desired them. In the next section we describe how we developed and administered our questionnaire about civic life on the Internet, and how our survey instrument produced some unintended consequences that both enriched and diminished aspects of our study. ii Methodology and its Consequences: Even though we brought considerable experience in questionnaire construction and survey research to our task, we soon discovered that posting a survey on the Internet differs from conducting conventional surveys in person, via telephone, or by ordinary mail. We encountered problems that in many cases neither we nor those we consulted had anticipated. These involved sampling, questionnaire forms, means of posting, and means of reply. The sampling problem was immediately obvious. There is no comprehensive list of individuals who use the Internet, nor is there any certainty about how many users log on from any particular node. In February of this year we posed the problem of drawing a representative sample to various colleagues at the University of Cincinnati (UC) and elsewhere and to the public opinion research (POR) mailing list. The responses we received helped us to appreciate the complications but offered no comprehensive solutions. Complications stem not merely from individuals having multiple accounts at various nodes or multiple memberships in various Internet groups (something analogous to having multiple telephone lines), but also from the ability of "lurkers" to read and reply to messages posted for groups to which they may or may not formally subscribe. Moreover, with a few key strokes anyone who reads or receives a posting can forward it to any other user.1 Over and above the uncertain number of individual users, the virtual communities themselves change from day to day as new groups come into existence and old groups are modified or die. In the end, as we could not sample individuals, we decided to sample from two types of user communities: Usenet newsgroups and Listserve mailing lists. We also decided to sample from two strata: ostensibly political and non-political groups. We based these decisions upon a combination of sampling theory and convenience. First, we could obtain fairly comprehensive listings of newsgroups and mailing lists;2 second, the central limit theorem would allow us to have some confidence in the distribution of characteristics of groups in each stratum if we could sample approximately 30 (or more) groups of each type. Knowing that some groups and lists would be inactive or dead, and that others would be closed or otherwise moderated so as not to allow a survey to be posted, we selected a systematic sample of approximately 50 groups from each stratum using a random start. No substitutions or replacements were allowed. The sampling design seemed simple but straightforward. We would post the survey to the selected lists and groups in a manner that seemed analogous to mailing a questionnaire. Subscribers to newsgroups would receive only the descriptive title of the questionnaire; subscribers to mailing lists would receive the questionnaire directly. The discussion of Usenet "Netiquette" in the UC Manual, (Clark, 1994: 41-42), consultations with local computer gurus, and pretests of the questionnaire led us to expect some refusals from mailing list managers and some flames from users who would be irritated by an "off-topic" posting. We were hopeful nonetheless that our explanatory letter would encourage cooperation or at worst mollify users who thought our posting inappropriate. Again, we viewed potential flames as the electronic equivalents of irritated refusals to ordinary requests for potential respondents to fill out a questionnaire or agree to an interview by telephone or in person. As experienced survey researchers, we had grown some pretty thick skins regarding refusals. Following customary procedures for questionnaire development, we circulated drafts of the survey instrument to friends and colleagues in mid-April via the Internet. After several iterations, we posted a revised draft to the POR mailing list and to the alt.politics.datahighway newsgroup. By early July the final version of the survey was ready for posting. The instrument itself consisted of 33 questions, the first six of which differed slightly, depending upon whether the survey was intended for subscribers to a mailing list or a newsgroup. The questions asked about the respondents' general experience using the Internet, their motivations for subscribing to newsgroups or mailing lists, and their experience regarding politically relevant activities associated with the Internet. These included requests from newsgroups or mailing lists for users to contact public officials, sign petitions, or otherwise become engaged in civic life.3 The entire document including the explanatory cover letter, took up just under 15K bytes as Microsoft Word text on a floppy disk formatted for the Macintosh. Pretests showed that it took the average respondent less than 15 minutes to fill out the questionnaire. Respondents were asked to return the completed surveys to the e-mail address of a special research account that the Academic Information Services unit of the Center for Information Technology Services permitted us to establish on UC's DEC-VAX cluster.4 Filling out a questionnaire electronically is not the same as filling out a questionnaire with paper and pencil or being interviewed face to face or by telephone. For many users it is difficult to skip forward and backward through a document. Had we enough time and money, we would have prepared a program that presented each question separately and provided the respondent with the appropriate pattern of skips and contingency questions based upon his or her previous answer(s). Instead, we settled for an ASCII text questionnaire designed so that the respondent could mark close-ended questions and fill in short answers using his or her newsreader or some other text editor. Feedback from pretests showed that to facilitate responses, space for answers should be left- justified, and lines should be under 72 characters.5 With our survey ready to go, we experienced an excitement tinged with hubris. While we knew we were not the first to sample users of the Internet, we saw ourselves as ahead of most social scientists, pioneers in adapting old techniques to new media. We hoped to be among the first to produce reliable estimates of selected political behaviors and socio-economic characteristics of those who use the Internet. We promptly made our first major technological error. Our pretests suggested that posting the questionnaire directly to subscribers of mailing lists would thrust a possibly unwanted-- and for many, an apparently irrelevant--instrument into mail boxes. Far better to seek the cooperation of list owners or managers, in much the same way that researchers seek the cooperation of officers of organizations they seek to survey. A notice to the managers would help establish our bona fides, and notices from the managers to subscribers would increase the response rates. Unfortunately, we had only a partial list of owners or managers. To obtain information about lists, our manual suggested issuing a List Detail and/or List Short command. These would return information about the selected lists, which we expected would include the e-mail address of the list managers. We knew that requests to subscribe should be sent to List Servers, not to list addresses. We erroneously assumed that the Listserve software would recognize the format of other commands (no subject, one line for each command). To our chagrin, our two line message was delivered to thousands of puzzled subscribers to unmoderated mailing lists. Most ignored the message: erroneous messages, particularly misdirected commands to subscribe or unsubscribe, are common. The documentation, when you can find it, is not easy to understand. (See Clark, 1994: 57-62.) The two line error would have been less problematic had it not been contained an indication that the message had been sent to multiple addresses: all the lists for which we did not know the managers. We received a few responses that expressed puzzlement at our sending apparently sophisticated commands improperly, and a few flames expressing dismay at our multiple listing. By responding quickly and politely we not only remedied the situation but we learned from our correspondents how to direct messages to list managers even when we don't know their names or e-mail addresses. Following these new instructions, which were not in our manual, we encountered no other serious difficulties nor any accusations of breaches of etiquette in distributing the questionnaire through the mailing lists.6 Now, however, came technological error number two. As the multiple addressing had caused problems with the mailing lists, we thought it would make sense to post our survey to each Usenet newsgroup separately. Like generals preparing for the last war, we reasoned that individual postings would show that we were not plastering (spamming) the Internet with our survey indiscriminately. Besides, as the survey resides on a board instead of being delivered to each subscriber's mailbox, it had less of chance of creating a disturbance. Indeed we had received no complaints about posting to Usenet groups in our pretests. How wrong we were! The Usenet software, in contrast to the Listserve software, is designed to handle multiple mailings as crosspostings. A crossposting resides on one bulletin board only; the crosspostings on other boards simply fetch the single posting when a subscriber calls for it. By approaching each board separately, we had posted 80 times more material than necessary. We didn't know it, but our posting caused considerable consternation and controversy on the Internet. Even with its inefficient use of space, our entire posting on the Internet took up less than one high density floppy, dished out in 15K bytes.7 At first blush, this does not seem like much, but it produced an unanticipated multiplier effect. Apparently, the unorthodoxy (bad Netiquette?) of sampling groups at random, as opposed to sampling only those groups whose titles suggested direct relevance to aspects of civic life, caused great controversy on numerous boards. Meta-discussions began, and these took up additional space on many boards.8 Such meta-discussions have an appropriate place on specialized boards, such as news.admin.policy, news.admin.misc or alt.current- events.net-abuse, but when they take place on other boards, they are viewed as using up precious disk space for "off-topic" subjects. While the controversy raged in the background, users were responding in droves. Although we received a number of the anticipated flames accusing us of everything from spamming to shilling for commercial enterprises, not to mention ignorance and net abuse, the large bulk of responses were completed interviews. A number of the respondents appended encouraging comments, and most requested to receive a copy of our report by e-mail. As we struggled to download and record the responses to make room for new ones, we remained blithely unaware of the magnitude of the controversy we had caused. "If your message is *really* important," reads a portion of news. announce.newusers, "pick up the phone and try to call the other person." The surveys had been posted on Tuesday and Wednesday, July 5-6. On Friday, July 8 we received a telephone call from a designated spokesperson informing us of the controversy we had caused. Indeed, our caller reported that resolving the controversy had become particularly difficult because our survey looked bona fide. Our caller then announced something to this effect: "I guess you don't know as yet, but 'someone' (sic!) has issued a cancel order to remove your survey from most of the Usenet boards." After much debate, the anonymous powers that be had decided that the survey could remain on boards that they thought appropriate, but they had ruled that our survey, however well-intentioned, must be treated as a spam.9 "I nearly made the mistake you made with a survey last year," our caller continued, "but I checked with our local news administrator, and he told me what would happen. Did you check with your news administrator before posting?" "News administrator? No. I've heard of such a person, but I thought that the administrator just decided which of the thousands of Usenet newsgroups the local installation would carry for posting. Sort of a grand censor, mostly to weed out boards devoted to pornography and the like." The UC manual makes no mention of the news administrator. We went on to discuss ways to post the survey that were less obtrusive. Crossposting to limited sets of newsgroups serially for specific periods of time seemed the most satisfactory substitute for our general posting. Other techniques, such as posting "pointers" inviting potential respondents to send for the survey, posting the survey on alt.usenet.surveys and asking respondents to go there to fill it out, or limiting the survey to "appropriate" boards seemed unsatisfactory. Our caller promised to send us information on some 360 academic surveys he had monitored on the Internet. Most were conducted by students, often with all the attendant difficulties of student efforts; nearly two-thirds had been limited to one or two newsgroups; none had been as ambitious as ours regarding its sample. The powers that be could not distinguish our method of sampling from what they define as a spam--an indiscriminate distribution of irrelevant messages, often for purposes, such as commercial gain, that are considered antithetical or irrelevant to the purposes of the Usenet newsgroups. To quote from e-mail downloads we later received, "Given the current controversy regarding spamming, their method of delivering the survey does not show maliciousness but rather a misguided enthusiasm and a lack of discussion with their newsadmin and colleagues." "You really should check with your news administrator," our caller concluded. To repeat: we had discussed the survey with our colleagues, both at UC and elsewhere. But as none had conducted surveys on the Internet, they, like us, had not anticipated the problems. As it turned out, we actually had discussed the survey with our news administrator. Indeed, he was the very person who had assigned us our special research account for the express purpose of conducting the survey! We only discovered this when we asked him if he knew our local administrator. He confessed to having so many other duties that he could devote only minimal time to news administration or guidance.10 He allowed that he had been receiving some e-mail about our survey, however. While the cancellation, coupled with the independent distribution of the survey (described in footnote 1 above), hamper our ability to claim "true" representativeness of the Usenet newsgroups that responded, we believe it had limited impact on the character or number of responses. Most groups have a default limit of seven to 21 days on postings. Active boards that we sampled, such as talk.politics.guns or soc.culture.malaysia, have even shorter limits. Three quarters of our respondents reported they checked their boards at least daily; fewer than 10 percent reported checking only weekly or less frequently. In short, events move rapidly in cyberspace. Users appear to respond to a message or posting immediately or else they delete or otherwise ignore it. Thus, we feel confident that our respondents, despite their self-selection, represent a sample of the most active users of the newsgroups and mailing lists. In any case responses trickled down after July 8, even from mailing lists that the cancellation did not affect.11 At this stage of our data analysis, however, we are less concerned with the representativeness of the data than with the problems of cleaning and verifying them sufficiently to allow us to test hypotheses suggested by the patterns of attitudes and behaviors discussed in section one. Because of complications arising from respondents who added categories, skipped questions, or otherwise introduced unreliability into the coding process we have limited our data analysis for this paper largely to describing the univariate distributions of relevant attitudes, behaviors, and socio-economic characteristics of our respondents. Finally, we discovered that in contrast to most refusals for telephone interviews or mail questionnaires, refusals for Internet surveys can be informative. Besides receiving 453 usable questionnaires, we received nearly 100 refusals. Of these 52 cited technical reasons: anything from improper posting methods to objectionable items in the questionnaire.12 Another 44 cited the irrelevance of questions of civic life or politics for their newsgroup or mailing list. They wanted to pursue the special interest to which the board or list was devoted, and they did not want to be bothered with other subjects.13 (We also received over 30 messages of encouragement, interest, and desire to help from respondents who returned questionnaires.) In the next section we present a first look at the responses to our questionnaire. iii A Profile of Internet Users: The Internet may be a world-wide web, but the 453 who returned completed surveys were hardly a cross section of the world population, of the population of western industrial states, or even of the United States. They were predominately male (nearly 80 percent), white (again nearly 80 percent), and young (median age of 31 years ).14 About 40 percent classified themselves as single and never been married. Asians, though underrepresented in terms of world population, outnumbered blacks and Hispanics by a ratio of 2 to 1 (4 percent versus 2 percent each). As a group, respondents were highly educated: 85 percent claimed to have some post-secondary (college or university) education; 51 percent reported having completed a degree program; and 29 percent reported having completed a post-graduate program. The respondents appeared to be affluent as well as educated. Approximately 44 percent reported living in a suburban location, and 38 percent classified their residence as urban. In contrast, only about 10 percent described their residence as rural. Less than 2 percent reported being laid off or unemployed, 2 percent said they worked only part-time, and only one percent were retired. Sixteen percent classified themselves as students.15 Nearly all the rest who reported their employment status claimed to be working full-time.16 The median household income of the 310 U.S. citizens who responded was between $40,000 and $59,000 annually. Over 22 percent of these respondents claimed an annual household income that exceeded $80,000.17 Despite their youth and affluence, however, the United States citizens who responded showed a distribution of party identifications that broadly reflects the general population: 36 percent Democrat, 32 percent independent and 23 percent Republican. (See Miller and Traugott, 1989: 81.) Nine percent, however, claimed some other affiliation. When asked to pick as many ideological descriptives from a list to characterize their "outlook toward politics and public affairs," as seemed applicable, respondents chose liberal (34 percent) and environmentalist (29 percent) most frequently. Moreover, they selected left wing (21 percent) and libertarian (20 percent) over right wing (6 percent). Twenty-three percent selected middle of the road, and 21 percent claimed to be conservative. Eleven percent described themselves as religious, 16 percent as feminist, 12 percent as Socialist, and 10 percent as indifferent or apolitical.18 As we would expect given self-selection, the respondents, on average, are quite active in cyberspace. Over half of the sample (approximately 53%) reported connecting to computer bulletin boards, Usenet groups or similar nodes in cyberspace at least daily, and nearly 60 percent of these reported connecting more than once daily (about 30% of all respondents). The median time spent weekly reading or replying to Bitnet and Internet e-mail or reading or responding to other information, programs or communications on computer bulletin boards, Usenet groups or the like was 5 hours. Despite their diligence in reading messages, however, only one third reported they replied to messages at least weekly or more often. About a quarter reported never having replied to a message until today (i.e., responding to our survey). Relatively few of our respondents reported they subscribe to any commercial on-line services: only about 25 percent. Compuserve and America ON-Line were the two most popular services (about 10 and 7 percent respectively) , followed by Prodigy (3 percent) , Genie and Dow-Jones (about one percent each). Four percent reported subscribing to a commercial service other than those mentioned above. As we noted, many Internet users, particularly those associated with educational institutions and government, have their usage fully subsidized. If Usenet and mailing list subscribers, on average, are more white, male, higher educated and more affluent than their compatriots, the relative infrequency of their subscribing to on-line services suggests that users of these services are perhaps an even more affluent and less representative socio-economic cross section of the general public. They, after all, are being billed directly for their usage of these services. Downloading information and receiving instruction, and obtaining others' information, ideas or arguments were cited most frequently by the respondents as to why they connected to bulletin boards, lists, or Usenet groups (about 74 percent each). In order of frequency respondents then cited entertainment and amusement (69 percent); professional/occupational contacts or interchange (61 percent); and expressing opinions; ideas and arguments (57 percent). Uploading information or instructing others and developing friendships and social contacts were cited with relatively less frequency: approximately 40 and 35 percent respectively. Lastly, only 17 percent reported they connected in order to organize groups or plan activities. Nonetheless, respondents reported a more active civic life in cyberspace than is typically reported by respondents in the national election studies (NES) of Center for Political Studies of the University of Michigan. Even though the technology is new, close to one-third had used e-mail to contact a public official. This compares to an estimated 28 percent of the NES who reported ever having written a letter to a public official during the 1960 and 70s. (Miller and Traugott, 1989: 295). About 60 percent had been asked to petition or otherwise contact a public official about an issue or public policy. iv Conclusions and Implications: Even though our data analysis remains in its early stages, we feel bold enough to venture two sets of tentative conclusions: the first concerns the advantages and disadvantages of several methods of using survey research on the Internet as a means to collect data about social attitudes and political behavior; the second concerns what the distribution of responses, including informative refusals and other communications, suggests about the extent to which civic life resembles the various typologies that were described in section one of this paper. The methodological complications that we discussed in section two highlight advantages and disadvantages of using the Internet to conduct survey research. While we could exploit the great power to distribute a questionnaire in ASCII format to a sample of subscribers to mailing lists and Usenet newsgroups cheaply and quickly, we soon discovered the virtual impossibility of assuring that those who responded to such a questionnaire were those for whom the questionnaire was intended. An ASCII questionnaire can easily be reposted to other individual users, mailing lists, and Usenet groups. For most newsgroups, it can also be filled out by non-subscribers who happen to browse the board. We know that reposting occurred. Thus, we probably would have experienced significant distortion of our original sample even if someone had not issued a spam-cancel order that removed the posting of the questionnaire from most Usenet boards. While distributing simple textual questionnaires to mailing lists and newsgroups can produce data suitable for exploratory analysis, development of more sophisticated sampling technique and survey instruments would seem necessary in order collect data suitable for testing formal hypotheses or measuring the fit of formal models. For mailing lists, the sampling problem can be approached by using a combination of political and technical strategies. Cooperation of the managers whose lists fall in sample can be solicited by e-mail, but also by telephone or "snail" mail if necessary. With the cooperation of managers and judicious use of the "Review" commands, thorough instructions can be attached to the questionnaire, specific individuals rather than whole lists can be sampled, and return rates can be measured.19 Even if managers will not cooperate, it is possible for researchers to subscribe to most lists in the sample, thereby acquiring posting privileges, at least for lists that are not strictly mediated even for subscribers. The sampling problem for Usenet groups is not so easily approached. If sufficient time is available, researchers could monitor each newsgroup selected for an extended period and then sample individuals who post or follow-up over that period. Following Netiquette, researchers may also negotiate with news administrators (via appropriate boards) concerning permission for posting "off topic" questionnaires for specified periods on selected newsgroup boards. Again , if time is not a critical variable, the postings can be limited to a few groups for short periods in order to lessen the chances of creating a controversy over spamming. If funding were available to acquire the necessary expertise, we would recommend developing a programmed questionnaire that would guide respondents through appropriate skip patterns, discourage or prevent them from entering inappropriate or uncodable answers, and facilitate feeding their coded answers into data analysis packages. It should also be possible to program the questionnaire so as to alert researchers if the respondent is not in the sample. In short, in order to assure sampling integrity it may be necessary to develop an e-mail analog of computer assisted telephone interviewing (CATI) as well as to negotiate with and obtain permissions from list managers and news administrators. All of this of course implies that democratic discourse on the Internet is not as open as it appears at first blush. The events that followed posting our questionnaire on Usenet boards suggest that we might liken civic life on the Internet to the interactions of the Greek citizenry with the behavior of their gods rather than to the interactions among citizens in the agora. Citizens may go about their business on the Internet, but in doing so they must avoid hubris lest they offend the gods. These gods, who control the destinies of citizens of cyberspace, rule by means of technological superiority rather than the exercise of immortal powers, but they rule nonetheless. They have the power to censure and to censor those who offend them, and some of them may even rain flame down upon those who violate their ritual Netiquette (Seabrook, 1994; Wiener, 1994). This is not to accuse the gods of having anything but benign intentions. We think that like the author of the informative refusal at the top of this paper, they seek to regulate behavior in the interest of preserving and protecting the integrity of the Internet. The Net is not infinitely expandable. It would become overloaded if too many newbies tried to exploit its one-to-many communicative powers. At the same time the gods need to encourage usage by a broad public if they are to justify continued public subsidies for the Net. Moreover, as access to the Internet becomes more common--some estimate there are as many as 30 million users--there are inevitably more sociopaths who have entered cyberspace. Currently, we find no evidence of systematic due process in the civic life of the Internet. The technological gods make the rules, codify them on various boards, and expect ordinary mortals to study these rules and learn the Netiquette before attempting any powerful ritual, such as invoking the Net's potent one-to-many communication capabilities.20 Although the gods may punish those who violate the Netiquette, they do not make the rituals easy to learn. Instructions on the proper invocation of commands are scattered among fugitive documents; not all commands are standardized; and not all of the gods give a high priority to their responsibilities for maintaining the Internet. The socio-economic characteristics of those who responded to our survey supports the interpretation that even ordinary users of the Internet represent a generally well-educated and affluent technological elite. These ordinary citizens have opportunities to improve their status. If they learn and practice the Netiquette, there appears to be nothing that forbids them from joining the technological deities. In this sense, the civic life of the technological elite of the Internet displays some democratic features. The second most important type of civic life seems to be that of the like-minded pursuing specialized interests. Next to technological complaints, respondents objected to our posting of a survey that did not concern the avowed special purpose of their mailing list or Usenet newsgroup. The idea of taking a random sample of groups was precisely to include avowedly non-political groups, but this idea does not correspond to commonly enforced rules of Netiquette. Other democratic features include participation in politics by contacting public policy-makers directly by e-mail or joining with others to petition them electronically. There is also evidence of a community of helpers who seem willing to share their time and expertise to help solve others' problems, provided that those who inquire have taken the time to locate and peruse the appropriate FAQs (frequently asked questions) before posting their messages. These latter types of civic life seem less commonly practiced, however, than the civic life of the technological elite and the special- interest users. Lastly, we did uncover evidence that users believe in the existence of a civic life controlled by those who manipulate virtual reality. A number of informative refusals alluded to the ability of users to disguise their true identities. Indeed, some accused us of disguising our true motives and sinister identities. We did not collect data on it, but we think that some of our critics in this regard may have squandered their civic energies on maintaining alternative identities in Multi-user Dungeons (MUDS). References Barber, Benjamin (1984). Strong Democracy: Participatory Politics for a New Age. Berkeley, CA: University of California Press. Clark, Mary ed. (1994). SAN Users Guide to the Internet: A Virtual World Traveler's Companion: for users of VMS and Ultrix. Cincinnati: University of Cincinnati Academic Information Technology Services (AITS) of the Center for Information Technology Services (CITS). Hardie, Edward T.L. and Neou, Vivian eds. (1994). Internet Mailing Lists, 1994 Edition. Englewood Cliffs, NJ: Prentice-Hall. Johnson, Otto, ed. (1993). Information Please Almanac, 46th ed. Boston: Houghton-Mifflin. Mansbridge, Jane (1980). Beyond Adversary Democracy. New York: Basic Books. Miller, Warren E. and Traugott, Santa A. (1989). American National Election Studies Data Sourcebook, 1952-1986. Cambridge, MA: Harvard University Press. Rheingold, Howard (1993). The Virtual Community: Homesteading on the Electronic Frontier. Reading, MA: Addison-Wesley Publishing Company. Seabrook, John (1994). "My First Flame." New Yorker, June 6, pp.70- 79. Wiener, Jon (1994). "Free Speech on the Internet." The Nation, June 13, pp.825-28. 1The problem is even more complicated. Some "list owners" or "list managers" oversee more than one mailing list. Several found our questionnaire of sufficient interest to offer to distribute it to subscribers on all their relevant lists. Meaning to be helpful, at least one apparently went ahead and forwarded it to these subscribers. Moreover, after the spam-cancel order, (see p. 8 below) we received a message from another user that he had posted the questionnaire on his FTP site and notified "five relevant newsgroups" of the controversy and of his new posting. 2To obtain the former, we sampled from a population of approximately 2,700 Usenet newsgroups that the UC makes available to users in the Newsrc file; to obtain the latter we sampled from approximately 4,300 Listserve mailing lists obtained from a list of lists on the Internet. (Both were obtained in early May.) These were supplemented by Internet lists in Hardie and Neou (1994) and lists recommended in communications from colleagues on the Internet and from the POR and PSRT mailing lists. We divided the newsgroups into political and non-political based upon their subject matter. (Rules for coding are available upon request). We divided the mailing lists in a similar manner, but we refined these classifications using information from Hardie and Neou (1994), the supplementary lists, and ultimately, the list owners themselves. 3To receive a copy of the questionnaire by e-mail, contact Margolis@UCBEH.SAN.UC.EDU. 4In retrospect, we should have established two accounts, one for political and one for non-political strata. Because of variations in the identifying information provided by the mailing systems of various computers that respondents used to return the questionnaires, it has often proven difficult or impossible to determine the particular mailing list or in some cases the particular Usenet newsgroup to which the respondent subscribes. At this stage of our data analysis, therefore, we cannot contrast the behaviors or characteristics of members of political versus non-political strata. 5Using ASCII text exacerbated another problem: many respondents, perhaps because they are accustomed to manipulating texts for posting or e-mail correspondence, chose to enter their own categories for close-ended questions. While our respondents' adding extra categories may provide some additional information, it has delayed the final cleaning and coding of the data. A programmed questionnaire could have prevented this or at least facilitated setting up codes for new categories. Such a questionnaire could also have been structured to feed directly into a data analysis program, such as SPSS. For this study, we printed copies of the returned questionnaires, adjusted the codes, and then used the SPSS data entry module to prepare the data for analysis. That is, we went from electronic medium to hard copy and then back to electronic medium, a process that introduced more opportunities for coding error than would programmed translation from returned questionnaire to SPSS system file. 6Some managers of course notified us that they chose not to distribute the questionnaire. The most common reason given was that the list was for a special purpose--usually a research project-- far removed, at least in the manager's mind, from anything to do with civic life or politics. We also discovered that mailing lists do not all share the same internal command structures for replying to messages. In some cases respondents who followed our instructions to "reply" returned their questionnaire to the list owner or, worse yet, to everyone on the list. Other mailing systems forced respondents to return the questionnaire in several parts. Still others were apparently so complicated that respondents e-mailed us a message expressing their inability to return the questionnaire. Two respondents (one in a pretest) returned the completed questionnaire by "snail" mail. 7Depending on the systems used by the boards on which it was posted, the survey occupied different amounts of space. The Listserve questionnaire with letters to the list managers and the respondents occupies 34 blocks on the University of Cincinnati's VAX. The Usenet questionnaire with its letter to respondents occupies 31 blocks. By comparison, the current Newsrc file, which contains the list of newsgroups automatically generated every time Usenet is accessed, occupies 227 blocks. 8An appropriate crossposting can direct all "follow-ups" to one board thereby reducing this multiplier. 9We later received records of some of the debate, which revealed the name of the "someone" who issued the fateful cancellation order. The cancellation was evidently issued on Thursday, July 7. 10Another rule of Netiquette, however, is "don't blame the news administrator for his/her users' actions." 11We received 483 responses of all types--comments, completed questionnaires, refusals etc.-- through July 7. We received an additional 67 on July 8. We received 101 responses thereafter, the latest arriving the fourth week of August. Messages from respondents and list owners informed us of one additional complication. Some Usenet newsgroups represent "bi-directional gateways" for certain mailing lists. Any posting on the newsgroup is automatically forwarded to all subscribers on the mailing list. As a result, the two populations of subscribers are not mutually exclusive, and some mailing list subscribers received copies of the Usenet version of our questionnaire. 12Although many of these refusals included helpful tips on how to improve our methods, they also included the most flames. Some threatened dire consequences for our sins, including sending us "mail bombs." We actually received one such bomb, which would have filled up our mailbox and thereby prevented questionnaires from being received, had we not discovered and removed it. We also discovered two cases where respondents filled out two copies of the questionnaire. To our knowledge no one deliberately stuffed our mailbox with duplicates or deliberately sent multiple returns from one account. Some of those who refused on technological grounds, however, claimed that each of these actions was easy to do. 13Several comments referred to our questionnaire violating the "charters" of their Usenet groups. A half dozen or so cited both technological and special interest objections. 14The age distribution was skewed positively: older persons tended to be outliers. Only 15 percent were born before 1950; over 25 percent were born after 1970. 15As the survey was administered during July and August when universities in most western industrialized countries are on summer schedules, we expect that this figure underestimates the proportion of students among subscribers to lists and newsgroups on the Internet. 16Missing data are a problem here. Nearly one in five respondents (18 percent) failed to report their employment status. Only 2 respondents classified their status as "keeping house." 17These figures compare with a median annual incomes of $37,000 for white households in 1990, and 13 percent of white households with incomes over $75,000 (Johnson, 1993: 46). 18These figures include both United States citizens and others. The distribution for the U.S. alone (N=305) is Liberal 36%; Environmentalist 31%; Libertarian 26%; Middle of the Road 26%; Conservative 25%; Feminist 20%; Left Wing 19%; Socialist 12%; Indifferent or apolitical 10%; and Right Wing 7%. 19Nevertheless, some subscribers have e-mail addresses that Review commands will not reveal. We expect that just as telephone subscribers with unlisted numbers have increased in number, mailing list subscribers with unlisted e-mail addresses will also become common. 20The gods also prefer to use the Internet as their primary means of communication. It is interesting that as the controversy over our posting raged, we received only one call from anyone who claimed some responsibility for regulation of the Internet. We received calls from a reporter for the Chronicle of Higher Education, however, and from an intern at the notorious law firm of Canter and Siegel.