Thus responses from the 2000 GSS are used in the regressions to represent the pre-9-11
period. The 2002 GSS as well as the combined 2002/2004 GSS are alternately used in the regressions
to represent the post-9-11 period. The 2002 and 2004 GSS were combined in some of the
regressions because the split-sample design of the GSS resulted in less data on the job satisfaction
question (satjob), the dependent variable, for these years. As Table 1 shows, there were n =
1,513 full-time workers who answered the GSS job satisfaction question in the 2000 GSS, but
there were only n = 722 full-time workers who answered the same question in the 2002 GSS, and
another n = 953 in the 2004 GSS. Thus combining the 2002 and 2004 samples provides more
balance in the pre- and post-9-11 sample sizes and also more statistical precision overall in the
regressions. It also provides a test of the sensitivity of the regression results to alternative definitions
of the post-9-11 period. For the descriptive graphs of the data, the GSS waves beyond 2004
are also included for purposes of illustrating the trend in job satisfaction in both sectors in the out
years (these sample sizes are also shown in Table 1).