Following training sessions conducted by RTI, Jackson Associates of Atlanta, Georgia, recruited survey participants in two large shopping malls in the Atlanta metropolitan area. Two malls were used to expedite the data collection. We chose
Atlanta because it is a large city outside the Central Flyway and is not located on a coast. Given the nonuse focus of the experiments, we wanted a low probability of recruiting users of migratory waterfowl in the Central Flyway. Although most of our respondents were users of coastal resources for recreation, the national policy context allows for a substantial nonuse component because it would protect all beaches, not just those used by the respondent.
Recruited participants were people over the age of 19 and were not full-time students. We established these screening protocols based on our pretesting experience. We decided to select adults who were in a position to make household decisions. Full-time college students also were eliminated because they often consider their parents’ income when making expenditures, and their WTP responses may not be true reflections of their own preferences and budget constraints. We did not give any participants an incentive payment.
Table 3-3 shows that we had about 400 completed questionnaires for each of the three migratory-waterfowl versions and about 400 completed questionnaires for each of the four oil-spills versions. The distribution of questionnaires and dates of administration for the two malls were:
• Southlake Mall—1,923 completed questionnaires between May 17 and June 5, 1991 • Lakeshore Mall—889 completed questionnaires between May 22 and June 5, 1991.
Table 3-3. Experimental Design and Number of Completed Questionnaires for Each Design Alternative 2,000 Birds 20,000 Birds 200,000 Birds Total Migratory Waterfowl: Open-ended format 398 408 399 1,205
Small Spills All Spills Total
Oil Spills: Open-ended format Dichotomous-choice format 406 393 411 397 817 790 Note: The goals for completed questionnaires were 400 for each open-ended alternative and 390 for each
Both Southlake and Lakeshore Malls are large, suburban, enclosed malls. Southlake Mall is located in the southern suburbs of Atlanta. It has 120 stores with four major anchor stores. Lakeshore is a smaller mall with two major anchors. It is located in Gainesville, Georgia, a more rural area north of Atlanta.
The recruiters used slightly different administration techniques at the two malls. In Southlake Mall, recruiters were stationed at one location on the upper level of the mall. The recruiters intercepted people and asked them if they would participate. If the respondent agreed, the respondent was escorted down a side corridor to a room and given the questionnaire. In Lakeshore Mall, the recruiters were allowed by the mall management to circulate throughout the mall to intercept respondents. Respondents then sat at benches in the mall and filled out the questionnaires using clipboards. On average, the surveys took 10 to 12 minutes to complete.
Table 3-4 summarizes the socioeconomic characteristics of the respondents for each mall. While most of the characteristics are quite similar across the malls, the Southlake Mall respondents are somewhat younger, have a higher income, and a much higher percentage of Southlake Mall respondents are males. Subsequent multivariate analysis indicated that the mall at which the respondent took the survey generally had no significant effect on the WTP responses, after controlling
Table 3-4. Characteristics of Respondents in Southlake and Lakeshore Malls
Southlake Mall Lakeshore Mall Mean Median Mean Median
Age (category in years) 30-39 20-29 30-39 30-39 Education (category) some college some college some college some college
Sex (% male) 65 NA 47 NA
Race (% non-white) 26 NA 22 NA
Income (category in $1,000) 35-50 35-50 25-35 25-35
People in Household (number) 2.9 3 3.2 3
Memberships in Environmental Organizations (number)
3 Mall intercept surveys have been deemed sufficiently reliable to be admitted as legal evidence (see McCarthy, 1991). The survey research literature states that an experiment needs only to be internally valid as long as the results will not be generalized to a larger population. Internal validity can be achieved sampling from any population when the sampling is random and the experimental and control groups are comparable. (See Sellitz, et al., 1970; and Babbie, 1979.)
for socioeconomic differences. These results are documented in Chapter 5, when we report the full empirical analysis.
We chose the mall-intercept approach as a cost-effective way to test our
hypotheses on people who could potentially be chosen for any CV survey. Our main goal is to have comparable samples for testing because the experiments are designed to compare differences across experimental treatments. They are not designed to develop damage estimates that are generalized to a specific population. Viscusi and O’Connor (1984) and Viscusi, Magat, and Huber (1985) have used mall surveys to conduct experimental tests in their risk studies. Additionally, mall respondents represent a broader group of the population compared with student populations, which have been used in several methodological studies (see Kealy, Dovidio, and Rockel, 1988; and Bergstrom, Stoll, and Randall, 1989).3
As discussed, each respondent received a randomly assigned survey version. This random assignment, coupled with the large sample sizes should result in subsamples with very comparable characteristics. Table 3-5 shows the distribution of sociodemographic characteristics by survey version. As expected, the distributions are very similar.