This survey differs from the other market surveys in that while they tried to map out the entire population and its propensity to use the Internet, for the Berkeley study only a specific segment of the population was observed: students [primarily undergraduates] currently attending a University.

The survey was administered to a large music class at UCB. The class was chosen because of the variety of ages and disciplines it has historically attracted. By and large the respondents were all undergraduates, although there were a few gradaute students as well. They were included in the analysis because there is little functional difference in their experiences and those of other undergraduates who had studied elsewhere before attending UCB. However, people who were over 29 were removed to consolidate the sample and because it is reasonable to believe that the experiences of a 31-year old are considerably different from those of a 21-year old. This survey deliberately had focused on a younger population because of the general assumption that of all Internet users at large, the greatest proportion of them are young and university-educated [Yankelovich, 1995-6; Hoffman, 1996]. Here then was a sample of people who were young and university-educated, yet it appeared unlikely at the outset that all of them had fully adopted the Internet. The survey was then designed to map out what within this young, university sample was shaping their usage behavior.

Approximately 250 surveys were distributed, of which 213 were returned. Of these, 204 were used following adjustments for age. Pertaining to sex, age, economic background, and race the sample was comprised as such:

• Sex: 93 people were males and 109 were females [approx. 46 and 54 percent, respectively, with 2 cases missing].⁸
• Age: Years of birth were adjusted to include only those from 1967 through 1979. Ages were consolidated into categories spanning two years, i.e., 1970 and 71 except for the first category which spanned from 1967 through 1969. The breakdown was

  

• Economic background: Respondents were asked what their family income was while they were in high school. The five original categories were compressed for analysis into three categories: $0-25,000, $26-75,000, and $76,000 and above.⁹ The breakdown was

  

• Race: This question was phrased, "When others are asked to describe you, what term might they use?" The choices were Black, White, Asian, Indian, Latino, and Other. The wording was chosen in part because there are too many ethnic possibilities to code efficiently and because if the economic standing of the individual or educational opportunities would have been affected by racial bias, then it was important to ask the question in those terms. For the analysis the "Others" were coded as missing, and Black and Latino were combined, as was Asian and Indian. The breakdown was

  

  Certain questions were designed to assess usage patterns. Question 23 asked which Internet software applications were used, and question 24 listed various possible uses for the Internet and asked how often the respondent used the Internet for any of them. For the analysis, the sub-questions to question 24 were divided up into 3 categories: Internet uses designed to take advantage of it as an information resource¹⁰, Internet uses designed to take advantage of it as an inter-personal communications medium¹¹, and the remaining questions referred to specific functions which could now be done on line¹².

  The rest of the questions inquired about what types of direct influences caused people to start using the Internet. These influences included jobs, high school classes, university classes, family, and friends. They differ from personal attributes such as economic background, gender, or race in that those were innate to the individual while the others were more recent external pressures that caused the individual to use, or perhaps avoid using, the Internet.

  Of the 204 respondents there were very few who answered the survey perfectly without misunderstanding or erroneously skipping any questions. There were some questions which required only one answer, so cases which had more than one response became coded as missing. While doing cross-tabulations, if the variable [question] being cross-tabulated with was coded missing for a particular case, than the other variable for that case became coded as missing and not counted for the cross-tabulation.

  However, while there were never 204 perfect cases for any question, there were rarely fewer than 150 usable cases either. The only occasions where the numbers were too small to work with were for some questions with many responses. In most of these cases responses were combined, reasonably, to generate more usable numbers. Some combinations were previously described for age, income, and race. The responses for question 16, which asked how often the Internet was used, were combined for some cross-tabulations to indicate the Non-Users ["Never Used" and "Used Once or Twice"] from the Internet Users ["Beginning to Use More" and "Frequent User"]. The four separate categories were used only when it was necessary to differentiate intensity of use.¹³

  However, there were still some groups which had considerably smaller populations to work with, such with the Black/Latino category which had fewer than 15 cases at any time to work with. The effect of the smallness of the data is that some results might be due to chance since the standard error was relatively large (consuming entire cases instead of fractions of them). This problem was generally overcome by graphing the percentages within each group, as opposed to the number of cases in each group¹⁴, and by avoiding deriving significance from smaller variations.

  Another coding difficulty arose from having offered various questions for free-response. The consequence of this practice meant it was often difficult to categorize and quantify the results. However, for both difficulties it should be remembered that the function of this study was not to definitively define patterns and tendencies but rather to map out those influences which might previously have been overlooked or over-simplified in order to suggest valuable areas for subsequent studies to look into.

  ---

  Back to Contents
  Next Section

  ---

  By Cathy Gellis, © 1996, 1998
  cathyg@csua.berkeley.edu
  http://www.csua.berkeley.edu/~cathyg