Survey Methodology and Data Reliability
The 1995 National Survey of Recent College Graduates (NSRCG:95) is sponsored by the National Science Foundation (NSF), Division of Science Resources Studies (SRS). The NSRCG is one of three data collections covering personnel in science and engineering, which constitute the NSF's Scientists and Engineers Statistical Data System (SESTAT). Further information about the design, implementation, and results of the NSRCG:95 can be found in the 1995 National Survey of Recent College Graduates Methodology Report.
The NSRCG used a two-stage sample design. In the first stage, a stratified nationally representative sample of 275 institutions was selected with probability proportional to size. Each sampled institution was asked to provide lists of graduates for sampling. The second stage of the sampling process involved selecting graduates within the sampled institutions by cohort. Eligible graduates were those who received bachelor's or master's degrees in the sciences and engineering from July 1992 through June 1994. Oversampling was employed to improve estimates for black, Hispanic, and Native American graduates. The overall sample size of graduates was 21,000.
The unweighted response rate for institutions was 97 percent, and the unweighted response rate for graduates was 86 percent. The weighted response rates were 97 and 83 percent, respectively. Thus, the net weighted response rate for the 1995 NSRCG was 81 percent, the product of rates at each stage of data collection. Interviews were completed for 16,340 graduates. The NSRCG:95 data were weighted to produce national estimates. The item nonresponse for this study was very low (typically about 1 percent) due to the use of CATI technology for data collection and data retrieval techniques for missing key items. However, imputation for item nonresponse was performed using a "hot-deck" method.
Different S&E fields were sampled at different rates, so weights were used to provide nationally representative estimates. The weights accounted both for the probability of selection and for survey nonresponse.
Standard errors for the survey were computed using a replication method known as jackknife replication. Tests of significance used in the analysis were based on Student's t. A Bonferroni adjustment was used to correct significance tests for multiple comparisons. The adjustment varied depending the on the number of multiple comparisons involved (i.e., the number of categories in the specific questions examined, and the nature of the hypothesis being tested). Statements of differences in the text are significant at the 95 percent confidence level after the Bonferroni adjustment.