Table 4.
Discrete choice experiment/conjoint analysis
| Paper | Scenario description | Question asked to respondents | Attributes | Econometric model(s) | Covariate results |
|---|---|---|---|---|---|
| Crime | |||||
| Iraguen and de Dios Orutzar | Respondents are asked to imagine they are traveling to work from home. The trip takes place on a regular working day, they arrive at their destination at around 7:45 am, and they drive their own car and are responsible for all costs involved | Respondents are asked to choose between two different routes with differing attributes | Travel time. Travel cost. Number of fatal car accidents per year | Multinomial logit model | Income negatively affects the perception of the importance of travel cost. Safety valuation is positively affected if the individual travels with somebody else. This is also true if the respondent is female or they have been in a serious accident before |
| Environment | |||||
| Dickinson and Paskewitz | Respondents are informed of the multiple types of mosquitoes in Madison (some a nuisance, some transmit West Nile virus). Control program which would control mosquito larvae could control one type of mosquito larvae or both | Asked to choose between pairs of hypothetical control programmes | West Nile Risk. Type of mosquito targeted. Cost (through taxes): $10–200 | Conditional logit model | Increased risk level leads to higher WTP. WTP decreases as cost increases |
| Flügel et al. | Respondents who had a recent trip by car were presented with different choices for a car trip route | Respondents are asked to choose between two routes with four differing attributes, 6 times | Cost: fuel and toll. Travel Time. Casualties: fatalities and serious injuries. Landslides: share of the route with landslide risk | Mixed logit models | Men are less likely to choose lower landslide risk. People with a higher education tend to choose the option with the lowest risk more often |
| Mattea et al. | No scenario given | Respondents are given six choice sets of seven alternatives, each of which consists of five attributes. These questions are asked twice, and respondents are given visual information on the possible options before being asked a second time | Four alternatives represent devices to protect against landslides: diverging channel, retaining basin, Video cameras, and Acoustic sensors. The fifth alternative is a hypothetical road toll | Mixed logit in WTP spaces. Multinomial logit model. Mixed logit in preference space | The ‘status quo’ is negatively perceived |
| Viscusi | Traffic—On an average day 100 people die due to traffic accidents. These risks are isolated deaths. Natural disasters—these are national catastrophes and large numbers of people can die at the same time. Hurricane Katrina killed over 1000 people. Terrorism—attacks by terrorists can also be catastrophes. The 9/11 attacks killed 2976 people | Respondents are asked ‘risk–risk’ tradeoff questions (traffic accident-terrorist attack, traffic accident-natural disaster) in two sets of 6 question blocks | Type of deaths prevented. Average number of deaths prevented | Conditional logit models. Mixed logit models | More education raises the utility coefficient in every instance, and more so with terrorism. Income has a negative effect on utility. Seatbelt usage increases the utility of reducing all deaths |
| Food safety | |||||
| Enneking | Participants are given a short introduction to the quality and safety labelling system (regarding liver sausages) | Asked to name three choices from a set of 6 sausages | Brand A: national premium brand (with/without Q&S label). Brand B: national brand (with/without label). Brand C: national premium brand—reduced fat (no label). Brand D: private label—organic (no label). Brand E: national organic umbrella brand name (no label). Brand F: private label | Maximum likelihood | Those who find low prices important avoid the more expensive labelled brands |
| Loureiro and Umberger | No scenario given | Asked to choose between two steaks (option A and option B) with five varying attributes | Price ($/lb). Country of origin labeled. Traceability to the farm. Food safety inspected. Guaranteed tender | Multinomial conditional logit model | Increasing price of option leads to lower utility. Steaks inspected by US food inspectors carry the highest premium |
| Smith et al. | No scenario given | Regarding improvement of food safety respondents are asked to choose between: the ‘status quo’, ‘Hire more inspectors’ and ‘purchase medicine’. Each subject is asked 12 choice questions, where each option consists of five attributes | Annual risk of food borne illness. Average amount of time you will be sick. Extra time needed to prepare food. Cost. Annual increase in income tax | Multinomial logit models | Consumers prefer reduction ex ante risk than ex post. Those who are more willing to accept risk, are not as likely to accept risk reduction policies. Respondents prefer private control over the risk reduction |
| Health | |||||
| Determann et al. | Respondents are presented with some combination of two scenario variables (1) susceptibility to the disease (2) severity of the disease. | Asked to choose between Vaccine A, Vaccine B and ‘No Vaccine’ in 16 choice sets. Vaccines are comprised of different levels of 5 attributes. | Effectiveness of vaccine. Safety of the vaccine. Advice regarding the vaccine. Media coverage. Out-of-pocket costs. | Latent class model. | Females and individuals who stated they would never get vaccinated were more influenced by media and more sensitive to costs. WTP is higher for more effective vaccines, especially if the outbreak was more serious. |
| Transport | |||||
| Patil et al. | No scenario given | Each respondent answered five choice exercises regarding their security preferences when traveling by train or metro | Type of CCTV cameras. How long CCTV information is stored. Who can access CCTV information. Security personnel at the station. Type of security checks at the station. Time to go through security checks. Security surcharge | Multinomial logit model | All preferred CCTV over no CCTV. Preference is weaker for younger people. Females have a stronger preference for CCTV |
| Rizzi and Ortuzar | Survey is disguised as a survey to improve interurban route policy and road safety. Respondents are given an identical trip in which: they drive their own car, they pay for the total cost of the trip, and they have to return after 20:00 | Respondents are asked to answer nine choice situations. They are asked to choose between two routes with differences in the three attributes | Travel time. Toll charge. Annual accident rate (represents “general level of safety”) | Binary logit models | Women have a higher preference for safety than men, as do older people. There is a higher preference for safety if the trip takes place at night. A person driving with others in the car is more aware of risk |