Experimental Design
Our experiment is a discrete choice experiment in a representative offline sampled online panel of the German population (Payback Panel). Each survey participant is presented with seven randomly selected scenarios from a series of 6435 possible scenarios, where he or she has to choose from two job offers in hypothetical cities. These scenarios differ in 7 attributes that can have three possible levels each. The first attribute is the hypothetical income of the respondents, which can be either the last monthly wage or an increase of 5% or 10%, depending on the scenario. The other six attributes are measures for the quality of different urban amenities and can take three different levels ("low quality", "medium quality" and "high quality"). These individual amenities are cultural offerings, social diversity, ecological quality, quality of the infrastructure, economic dynamism and family friendliness of a city.
Our design of the choice scenarios is based on a resolution 4 orthogonal array (L243.3.20) of the possible attribute combinations, which contains 243 alternatives. We choose this design array as a starting point so that we can estimate the effect of two-level interactions of attributes. If all possible unique combinations of the 243 alternatives are considered, there would be 29,403 possible decision scenarios.
However we impose three different restrictions on the design. First, given that 6 of our attributes are either the respondents wage or a public amenity that has a clear quality ordering, we exclude strictly dominated choices from the set of decision scenarios. Second, for scenarios, where the wage is equal in both alternatives, we only keep the ones where this is the last monthly wage. Third, to reduce the cognitive load of the scenarios we only retain choices where a maximum of 3-4 factors are different in the two alternatives.