Survey methods used in 2006

The survey was administered by the Institute for Social Science Research at Arizona State University from April 29 through September 27, 2006. Data was collected using a multi-modal approach. Below is an outline of the sampling and survey process.

Sample selection

40 neighborhoods selected; n=20 in each neighborhood. GIS (Geographic Information Systems) created maps of each neighborhood from the Maricopa County Tax Assessor’s records with each parcel of land noted with an APN (parcel ID) number . Each neighborhood was canvassed to determine the accuracy of the maps. Any discrepancies were noted. (For example, a house may be listed on the map but may have been torn down.). All business were noted on the maps. All apartment buildings were noted on the maps. The number of apartments in each apartment complex and the apartment numbering were noted. For each neighborhood, a list of eligible addresses was created from the maps and APN numbers including apartments, mobile home parks and so forth. Using SPSS, 60 addresses were randomly selected from each neighborhood. The first 40 addresses were used as the sample. These addresses will be referred to as "sample". The other 20 addresses were placed in reserve to use as replacement sample (primarily when a home was vacant) or to use if additional sample was released.

Pre-survey efforts

Doorknob hangers announcing that PASS was coming were placed on the doorknobs of the sample. . The following was sent to each sample: A letter explaining the project and how to complete the survey in English and Spanish;. A brochure with "fast facts" about the project in English and Spanish;. A $1 bill; and,. A magnet that was graphically designed with PASS printed on it. The above information (in "3") was sent out over a period of approximately 3 weeks. Information was sent by neighborhood. A phone line dedicated to PASS was installed at the ISSR. Sample was asked to call the number where they would reach a message (in English and Spanish) and be asked to leave information and a time for us to call them back. Phone numbers were obtained for as many addresses as possible. Using www.whitepages.com, addresses were entered and then phone numbers were accessed when available. We were able to find phone number of approximately 50 percent of the sample. About 25 percent of those phone numbers were inaccurate. Few phone numbers were obtained for apartments.

Methods of completing the survey

Random selection within each household. In the letters sent to the households, the scripts interviewers read to adults who answered the phone in each household, and in instructions provided at the beginning of the web survey, households were requested to allow the head of household who most recently celebrated a birthday to complete the survey. Language. All correspondence was provided in English and Spanish;. All survey methods (online, phone and in-person) were provided in English and Spanish. Online: A simple URL address was provided where the survey could be accessed online. Each sample was provided a unique login ID and password that the respondent had to enter to log onto the survey. The login ID and password were connected to the address. When an individual accessed the website, s/he had to enter the logon ID and password. The computer called up with address associated with that login ID and password. The respondent had to confirm the address before proceeding. Telephone: Telephone interviewers called numbers provided through the phone number search (see "2e" above) and asked to conduct the interview at that time or to set up an appointment. Interviewers called each phone number a maximum of 20 times unless the respondent insisted s/he did not want to complete the survey. Interviewers also called appointments made through the hotline. In person: . Sample could request an in-person interview. Interviewers went to the addresses of households in the sample that had not completed the survey after the survey had been in the field for approximately 2.5 months. English and Spanish-speaking interviewers went to each sample address where no one had completed the survey and did one or more of the following: Conducted the interview at that time;. Interviewers took laptops with them so that they could access the questionnaire and record the responses. Left a flier (prepared in English and Spanish);. Provided a card with the URL and the sample’s login ID and password (in English and Spanish);. Obtained a phone number; and/or obtained an appointment for an interview. Follow up for non-response: . Letters and/or postcards were sent periodically to each address that had not completed the survey (except for individuals who specifically requested us not to contact them) asking them to take it, reminding them of how to take it (online, phone, etc.), and providing their login ID and password. Interviewers left messages on voice mail or answering machines asking sample to take the survey and providing basic information on the methods. Interviewers visited the addresses of non-respondents (see "3e" above). Incentive: Initially, a $20 incentive was offered for individuals who completed the survey. In mid-July, the incentive was increased to $50 to help obtain responses.

Response rate

1,584 pieces of valid sample were used. From the valid sample, 808 completed surveys were obtained resulting in a response rate of 51 percent. Additionally, sample usage was monitored by neighborhood. In every neighborhood, a minimum response rate of 50 percent was obtained. Of these 808 people, 59% took the survey on the Internet, 34% were surveyed on the phone, and 7% were interviewed in person. Respondents had the option of taking the survery in English or Spanish. It took an average of 45 minutes to complete the phone survey.

Phone and web surveys

Both of these modes used a software system called WinCati which is produced by Sawtooth Technologies.

Temperature

simulated temperatures that was created to examine temperature variability among 40 diverse neighborhoods studied as part of the Phoenix Area Social Survey (PASS) 2006. Susanne Grossman-Clarke ran the Weather Research and Forecast (WRF) meso-scale climate model for a 96-hour period (July 15-19, 2005) representing a four-day heat wave that occurred throughout the Phoenix metropolitan area. The output of the climate model was analyzed to compare exposure of threshold temperatures among the 40 neighborhoods within the study area.