38 This approach is explained by Harrison and Lesley [1996]. The alternative is to estimate the statistical model using information on the sample design, in order to generate estimates at that level reflecting the population. We plan to examine and compare both methods in subsequent work.39 It is possible to undertake experiments over the web with a large sample of subjects drawn from the population. Kapteyn and Teppa [2003] illustrate how one can elicit hypothetical responses to elicit time preferences using a panel of 2,000 Dutch households connected by home computer to surveys. Although not concerned with risk and time preferences directly, Hey [2002] illustrates how one can augment such electronic panel surveys with real experiments. Donkers and van Soest [1999] elicit hypothetical risk and time preferences from pre-existing panels of Dutch households being surveyed for other reasons.
đang được dịch, vui lòng đợi..