The effect of rest period on response likelihood

Given the multitude of survey invitations issued to an individual, survey researchers worry about the impact of repeated requests on participation in their specific studies. This analysis uses the extensive survey cooperation history of one government agency, Statistics Iceland, and complementary registry data to identify individuals selected into more than one sample for all Statistics Iceland general population household surveys over a 12-year period. Examining these 10,605 individuals, we model whether the decision to participate in the second survey depends on the length of time between survey exposures (i.e., rest period) and if this effect varies by attributes of the first survey for which the case was sampled. The results show a weak linear positive effect between the length of the rest period and participation in the second survey request, but this is explained by a combination of demographic characteristics and survey indicators. A persistent effect in the final model is the large, positive correlation between cooperation with the first survey and cooperation with the second. Also, being sampled first for a burdensome survey increases the likelihood of participation in a subsequent survey; the effect on nonrespondents in particular introduces an interesting twist to understanding how multiple survey invitations affect one another over time. Finally, being sampled for the same survey twice is detrimental to cooperation. While the objective of the study was to quantify the effect of rest period, the results more so highlight the importance of better understanding repeat respondents to improve protocols for gaining cooperation and address potential nonresponse bias.