Table 1 shows the DK levels for the 2002 GSS and for four different methods (or treatments) for handling DKs in the KNS. In the GSS, Don't Know was an unread response. The interviewer used a special key (F8) on his or her computer to record each DK response given by a respondent. The KNS versions were administered to random subsamples and consisted of the following formats:

• Treatment A (KNS-A)—An instruction on the first screen directed respondents to skip questions to indicate "Don't Know/No Opinion." It read, "In this survey we'd like your opinions about some important national issues. There is no right or wrong answer, and please answer the questions based on your knowledge. If you do not know the answer to a question, you can simply skip the question to indicate 'Don't know' or 'No opinion.’" Respondents were not offered a Don't Know option on the screen.

• Treatment B (KNS-B)—Don't Know was presented on the screen as a response option for each individual item.

• Treatment C (KNS-C)—Don't Know was presented on the screen as a response option for each individual item. Then, respondents who initially selected Don't Know were prompted to select a response on a follow-up screen that asked, "If you had to choose, are we spending too much money, too little money, or about the right amount of money on...?"

• Treatment D (KNS-D)—Instructions on all the screens told respondents, "If you do not know the answer to the question, you can simply skip the question to indicate 'Don't Know' or 'No Opinion.’" Items were skipped by clicking on a "Next Question" button. Don't Know was not listed among response options.

Table 1 shows statistically significant differences between the GSS and the KNS for almost all items on treatments B and D and for seven of the first eight items on treatment C. Treatment A shows only five statistically significant differences, all of them modest in size.

Table 1

Don't Know Levels in GSS and KNS

                                                         *p<.05 KN significantly different from GSS

Table 2 averages and summarizes results from Table 1. It shows that the Don't Know levels averaged the lowest for the KNS-A treatment (with skip-out instructions on the initial screen only) at 3.2 percent for the first eleven standard items, followed by the GSS at 3.7 percent. DK levels for KNS-C came in at 10.6 percent (with the probes converting 93 percent of all DKs to substantive responses, which were disproportionately “about the same”). Don't Know levels for KNS-D were at 11.0 percent and for KNS-B at 16.2 percent. Clearly, Don't Knows were mentioned more frequently when they were made more explicit and acceptable. This was consistent with the literature on format-variable DK levels.

Table 2

Average DK Levels by Mode, Format, and Question Order

This pattern also appeared within the GSS between 2000 and 2002, when the GSS shifted from paper-and-pencil interviewing (PAPI) to computer-assisted-personal interviewing (CAPI). The  percentage of respondents mentioning DKs declined in 2002, presumably because CAPI uses more of a skip-out approach while PAPI has DKs explicit among the response options. Across forty-nine opinion items, DKs averaged 4.4 percent on the 1998 and 2000 GSSs and 2.6 percent on the 2002 GSS. This indicates that interviewers as well as respondents were influenced by the format of DKs in questions.

Altogether, the implication of these findings is that a web format for Don't Knows—KNS-A—can be devised to produce results similar to those in face-to-face surveys such as the GSS.