The study population for this survey was all undergraduate alumni who received degrees from the University of Virginia in 2002, 2001, 1998, 1993, and 1983. A total of 4,780 alumni were selected according to a stratified random sampling procedure using undergraduate school as well as racial and ethnic backgrounds as the strata variables. This was done in order to ensure a sufficient number of respondents from the smaller schools (Architecture, Commerce, Education, and Nursing) as well as minority respondents, in particular African-, Hispanic-, and Native-Americans.

All respondents were sent personalized letters, signed by University President John Casteen, on September 22, 2003. The letter informed them of the importance and general purpose of the survey, assured them of confidentiality, and invited them to participate. Initially, the survey was conducted over the web. Respondents were tracked via an alphanumeric randomized respondent key. Five email reminders were sent to those alumni who had not responded on October 10, November 18, March 9, April 14, and May 5. Two postcard reminders were mailed on October 29 and April 19. A mailing that included a paper version of the survey was sent on February 9, 2004. Ninety-four percent of the responses were received over the web. Lotteries for $25 and $50 gift certificates from the University bookstore were conducted as an incentive for respondents.

One-hundred and nineteen respondents could not be contacted because of bad addresses, and eleven respondents requested removal from the survey. This brought the actual sample size down to 4,650. When the field period ended on May 31, 2004, 1,662 responses had been received, for a response rate of 36%. At a 95% confidence level, the overall sampling error for the survey is 2.2%. For the Tables of Results and subsequent analyses, we used two comparison groups-African-Americans formed one group and the second group combined white students with students from other racial groups. [1] The response rates, number in the sample, number of respondents, and sampling errors by the two comparison groups are reported below.