Web Appendix of "Ambiguity Attitudes in a Large Representative Sample"; Not for publication

To ensure that its sample is representative of the Dutch population, LISS uses careful procedures to randomly select households from the population register compiled by Statistics Netherlands. 1 Selected households are initially contacted by letter and then by telephone. LISS makes up to 15 attempts to contact a household by phone. If telephone recruitment fails, a LISS representative visits the household. If, after eight in-person visits LISS is unable to establish contact, a new household is selected. To limit the possibility of sample selection bias, recruited households are provided with free computers and internet access if necessary. To encourage participation and retention within the panel, subjects are paid for each survey they complete. Knoef and De Vos (2009) show that the LISS panel is generally representative of the Dutch population. See http://www.lissdata.nl/lissdata/ for more information. In January 2010 LISS fielded a survey module designed by the authors. From the full LISS panel, CentERdata randomly selected 2,491 individuals to participate in this survey, and 1,935 responded (77.7%). Of the 939 subjects in the real incentives group, we exclude 164 because of missing income or financial asset data, 61 for answering “Indifferent” to all ambiguity questions including the two check questions 2 , 38 for spending three seconds or less on each set of questions, and 10 for missing other variables. This leaves a sample of 666 subjects. We exclude the subjects who answer indifferent to all of the ambiguity questions or who spend three seconds or less on each of the question sets on the grounds that these subjects likely did not expend effort on their choices.