Statistics producers are increasingly using the web as a mode of data collection. At Statistics Sweden, the web is in self-administered social surveys increasingly offered as an alternative to ordinary mail questionnaires. The motives to offer the web alternative lie in quality and financial considerations such as automated online editing, reduced mailing costs and less scanning work.

Nevertheless, the use of the web as a response mode has been disappointingly low: usually, between 10 and 15 percent of the respondents are using it. A preliminary study of ours indicated that the respondents, when offered to participate in a self-administered survey, are inclined to select the mode that is immediately at hand when the request for participation arrives.

In an attempt to increase the proportion of web respondents, we conducted a large-scale experiment where we varied the delay of providing the paper version of the questionnaire to the respondents. In the standard approach, both the paper questionnaire and the login data for the web survey were included in the first mail-out. In the experiment groups (A1 to A4), the time delay of presenting the paper questionnaire to the respondents was varied, from no delay in A1, to three weeks after the initial mail-out in A4.

The results show that the standard approach had a proportion of web respondents of 14.5%. The proportion of web respondents in the experiment groups varied from 2.6% in A1, over 45.0% and 44.6% in A2 and A3 respectively, and up to 64.7% in A4, thus showing potential for considerable raise of proportion of web respondents.

The alternative approaches seem to have had a minor disturbing effect on the over-all response rates: the standard approach had a 75.7% response while the four experimental groups had 74.8%, 71.4%, 72.0% and 73.3% respectively in A1-A4. (Only two of the differences are statistically significant: S-A2 and S-A3.)

Calculations show that the alternatives A2-A4 could lead to a cost reduction of between 12% and 20% compared to the cost of the standard strategy.