PAYS is a longitudinal, dual-site study being conducted with seventh grade students and their families in two metropolitan areas; Phoenix, Arizona, by the Prevention Research Center at Arizona State University (ASU) and Riverside, California by the Department of Sociology at University of California, Riverside (UCR). The desired sample size was 200 families from each of the two sites, with 50 families per site in each of the 4 populations of interest: Anglo Intact families, Anglo Step-father families, Mexican-American Intact families, and Mexican-American Step-father families. “Intact family” is defined as the child living with both biological parents in the same household. For “step-father families”, the child needed to be living with his/her biological mother and man “acting in a father role” who was not the biological father; this household structure needed to be existing for one year or more. The mother and step-father did not need to be legally married. The ethnicity requirement was that all three family members participating in the study were required to be of the same ethnic origin, either Mexican or Anglo (white, of European descent.)
Below we summarize the PAYS sample by ethnicity and family type.
European American Intact families = 110
European Amerian Stepfather families = 89
Mexican American Intact families = 107
Mexican American Stepfather families = 86
Total = 392
The sample acquisition process was substantially different in the two sites, owing to the different rules and statutes governing such data gathering in the two states. Accordingly, documentation herein will be separated for the two.
The following describes the sampling, recruitment and data collection at the ASU site.
We used the metropolitan area schools as the starting point for acquisition of the sample. The ethnic make-up and enrollment statistics for all schools in the metropolitan area were examined and 10 schools that each had at least 30 percent Mexican-American students and had substantial enrollments were selected to be approached about participation. Study investigators designed a written project summary and mailed this to the Principals in each targeted school, following that contact with a personal meeting to further detail the project and elicit the school’s cooperation. Two schools declined participation; the other eight were highly enthusiastic and cooperative.
A very brief (4 item, 5-minute) postcard survey was administered in the seventh grade classrooms of each of the schools early in Spring semester, and repeated for a new cohort of seventh graders early in the following Fall semester. The questions asked only about the ethnicity of the child and parents, and which adults the child lived with. Since the child was being asked to divulge information about the family without parental informed consent (although two of the schools required a passive consent letter to be sent to parent’s prior to the administration of the postcard survey), PAYS staff could not administer nor observe these responses. Instead, with permission from the schools, we employed a bi-lingual member of each school’s personnel (known as the school recruiter, who, as a member of the school staff, had the right to collect this kind of data, so long as it was not divulged to the research team without explicit parental permission) to execute this phase of the data collection.
The postcard procedure was conducted in each school during one week in the second or third week of the Spring (Cohort 1) and Fall semesters (Cohort 2) in 2003. The goal was to conduct the survey in a class subject that reached all 7th graders yet did not duplicate administration to any child. School recruiters selected which subject classes to use to administer the survey. Classrooms varied between schools and included Homerooms, English/Language Arts, Science and Social Studies classrooms. Specific study explanation and instructions were sent to the classroom teacher along with pre-numbered survey postcards. Depending on the needs of each individual school and/or recruiter, either the classroom teacher or the school recruiter administered the survey.
Once the postcards were gathered, the school recruiter sorted the post-cards into five categories:
- Those that appeared to be eligible as Anglo intact families,
- Those that appeared to be eligible as Anglo step-father families,
- Those that appeared to be eligible as Mexican-American intact families,
- Those that appeared to be eligible as Mexican-American step-father families, and
- Those that fell into none of the above categories (e.g., single parent families, mixed ethnicity families, African-American families) and thus were not eligible to participate.
The results of the postcard survey procedure, responded to by 5,415 children over all 8 schools (ranging from 595 in the smallest to 1,019 in the largest school), showed that the rate of ineligibles (“none-of-the-aboves”) was 55% (with a range over schools of from 36% to 62%); the rate of Anglo intact families was 12% (with a range of from 1% to 26%); the rate of Anglo stepfamilies was 3% (with a range of from 1% to 4%); the rate of Mexican-American intact families was 25% (with a range of from 14% to 50%); and the rate Mexican-American stepfamilies was 6% (with a range of from 3% to 12%). In addition, 6% were either absent the day of the administration, answered that they didn’t know, or refused to answer. This led to the identification of 2,459 students who appeared eligible for the study, of which our desired sample size was 200, 50 per cell. Intact families were about 4 times as numerous as stepfamilies, for both Anglos and Mexican-Americans.
However, the number apparently eligible varied greatly over the 4 cells: 660 Anglo intact; 159 Anglo stepfamily; 1318 Mexican-American intact; and 322 Mexican-American stepfamilies appeared eligible based on the postcard survey responses. Accordingly we employed a randomization procedure to select from among the apparently eligible those we would attempt to contact and recruit into the study. Specifically, the school recruiters reported the postcard information for each student to the PAYS staff on tally forms that only listed the students’ identification number we had assigned, thereby keeping the child’s information secure and confidential. Then the office gave the school recruiters a list of a randomly selected subset of the ID numbers to attempt to contact during each week of a 10 week recruitment period until the targeted sample size for each category was acquired. The school recruiters extracted parents’ names, addresses and phone numbers from the school records for the children we indicated were to be randomly contacted that week and mailed them a brochure and a letter introducing the PAYS project to the family one week prior to the time the family was scheduled to be contacted by telephone. The brochure (see appendix A) emphasized the PAYS project’s benefit to the school and family, indicated that each family would receive a stipend of $120 for participating in the interviews ($40 each for mother, step/father and child) and encouraged them to call into the office for which they would receive an extra $5. For cohort 2, we increased the incentive amount to $200 for Mexican-American Step families.
For cohort 1, two different types of phone contact were conducted. 1) Four of the school recruiters who had been successful during a pilot study were trained to call the family and conduct the full recruitment procedure. This included questions about the family’s interest in participating and their eligibility through verification of the information that was on the post-card survey. Then they forwarded the recruitment forms to ASU, identifying the results of the phone contact and ID number if the family was unwilling or ineligible, and the eligibility data and contact information for willing families. 2) The other four recruiters conducted a “short recruit” asking the family for permission to have ASU contact them, without asking for eligibility information or willingness to participate. These families were told that they would receive $5 just for allowing ASU to contact them by phone. Once the family agreed to the call from the PAYS staff at ASU, a letter and $5 cash were sent to the home. Within one week of that mailing a recruiter from the project staff contacted the family and administered the screening and eligibility form.
In cohort 2, based on our better success with the short recruit procedure, we changed recruitment by having all school recruiters use the short recruit procedure as described above, asking families only if ASU could contact them, in order to have more control over the work flow and in an effort to increase response rates.
The school recruiters were trained in two three hour sessions. Session one was devoted to conducting the postcard survey, tallying the results and reporting back to the Project Director. The second session involved staging the mailing of a brochure and letter and calling the family. Intensive training on interviewing protocol and administration of the specific eligibility and screening form included role playing.
The phone and letter contact was eventually attempted with 640 families, in order to achieve the final AZ sample size, ranging from sampling 12% of the apparently eligible Mexican American intact families to sampling 93% of the apparently eligible Anglo stepfamilies.
After making letter contact with the 640 families, the ASU staff of recruiters began their telephone calls to recruit participation. It was made clear in this phone call that participation of mother, (step)father, and adolescent were all required, and that we would send a team of three interviewers to the family home. Results of this recruitment attempt are shown in Figure 1. As shown, we were successful in re-contacting 499 (77.9%). Of these, 124 (24.8%) were found, upon rechecking, to not meet our eligibility requirements, stated above, and another 133 (26.6%) refused to participate. We got agreement to participate from 242. Of these, 2 became unlocatable, 11 became ineligible, and another 25 decided subsequently to refuse participation; thus, we actually interviewed 204.
What is the response rate for our sample (the number of interviewed families divided by the number of eligible targeted subjects)? As discussed in Braver and Bay (1992), ineligibles cloud the issue, since eligibility “cannot be unambiguously decided without contacting the subjects, yet a certain unknown percentage of non-contacts would undoubtedly have been ineligible as well” (p. 928). They recommend reporting two response rates as upper and lower boundaries. One is a “conservative” estimate, which assumes none of the non-contacts would have been found ineligible. Based on this assumption, our response rate was 41.4% (204 interviewed divided by 640 attempted to contact less 12 who we later declined to interview since our quota for that cell was met, less 124 more originally known ineligibles less 11 more who became ineligible). The other, likely “liberal estimate” assumes the proportion of ineligibles among the non-contacts is the same as it was among those contacted. The percentage of known ineligibles among those we could contact to determine their eligibility (135 known ineligibles plus 204 interviewed plus 62 refusals) is 33%. Assuming 33% of the 225 we could not contact would also have been ineligible, this subtracts another 76 from the denominator, making our liberal estimated response rate 48.9%.
We used two local school districts as sources for recruiting the sample population, the Riverside Unified School District and San Bernardino City Unified School District. Each school district had an ethnic make-up of at least 30 percent Mexican American student enrollment at all of their middle schools. The Principal Investigators met with district assistant superintendents to summarize the project and explain project goals. The superintendents agreed to make the project a collaborative effort and allowed the principal investigators to correspond or meet with middle school principals. Six of six middle schools in the Riverside Unified School District agreed to participate in the study and six of eight middle schools in the San Bernardino City Unified School district agreed to participate (the two non-participating schools in San Bernardino had fewer Mexican American students and did not respond to initial inquiries).
The UCR PAYS team used two methods for initial recruitment of families: school recruiter telephone calls and mailings with postcard replies. Both methods relied initially on the analysis and sorting of school records. In the San Bernardino City Unified District, the school recruiters examined electronic emergency contact records. In the Riverside Unified District, the district office provided school enrollment data and contact information for the families as explained below. In both cases, the records contained seventh graders’ first and last names, as well as the first and last names of their parents. Based on the last names of the three target family members (seventh grader, mother, and father/stepfather) spreadsheets were created with a sorting of those initially believed to be in one of the four population groups (Mexican American Intact, Mexican American Step, European American Intact, European American Step). For example, families in which all three members had the same last names were tentatively classified as eligible for the intact family groups. Families in which the seventh grader and mother had the same last name and the father had a different last name were classified as tentatively eligible for the step family groups. Ethnicity was estimated using Spanish surnames. Following state regulations and local practice, PAYS staff were not allowed to telephone the families without their prior consent. Instead, school district staff members were employed by PAYS to make initial telephone calls to the families. When the school recruiters made contact with the families, they explained the PAYS research project and asked screening questions to confirm that the families met eligibility criteria . If eligibility and interest were confirmed, the family information was turned over to the UCR PAYS staff and after follow-up screening telephone calls were made, an interview was scheduled by UCR PAYS staff.
The second recruitment method, used for the Riverside District, was based upon similar preliminary spreadsheets created through district records intended to identify potential membership in each of the four recruitment cells. PAYS Staff conducted a mailing to the target population families containing three items: (1) A brochure explaining the project and the eligibility requirements; (2) A letter from the district superintendent announcing the project and the district’s collaboration, and (3) A self-addressed stamped postcard that the families could fill out with their names and phone numbers and return to the PAYS staff if they were interested in participating. Upon receiving the post card, PAYS Staff called the family to further explain the project and screen the family to confirm their eligibility. Once eligibility and interest were confirmed, the family was again contacted to schedule an interview. Both recruitment procedures were used during the Spring of 2003 (Cohort 1) and Fall of 2003 (Cohort 2).
State and local regulations prohibited in-school collection of initial family composition data from children as was done in Arizona, so the UCR sampling strategy is not directly comparable to the ASU sampling strategy. Because we relied on school recruiters and letters to make initial contact with families, and contactrf only those who expressed some interest in learning more about the the study, we do not know how many families were eligible to participate based on our initial criteria. Although we cannot compute a conventional response rate, we can estimate refusal and acceptance rates for those we contacted.
We estimate that 77-83% percent of the eligible families we contacted agreed to be interviewed. Of the 540 families who expressed interest to school recruiters or by returning the postcards, 66 (12%) had bad phone numbers or were not reachable. Of the 474 families that we were able to contact, 165 (35%) were ineligible according to our sampling criteria and another 61 families (13%) were not invited to be interviewed because we filled the respective cell before interviewing them. Of the remaining 248 families, 29 (12%) were determined to be eligible and refused to participate, another 27 (11%) with unknown eligibility refused to participate, and 192 (77%) were eligible and agreed to participate. We do not know exactly how many of the 27 refusers in this group met our sampling criteria, but we estimate that if a similar proportion of this group and of the total contacted were ineligible (45%) then our total refusal rate among eligible families was 17% (41/248). In contrast, assuming that all families in this group were eligible, the total refusal rate would be 23% (56/248). Refusal rates varied across family types, with stepfather families much more likely to refuse to participate in the study than birth-father families. The following table presents data on the number of families with no phone or bad phone numbers, the number contacted, their eligibility, and estimates of refusal rates and net response rates.
UCR Contacts, Eligibility, and Response Rates by Family Type
|Anglo Intact||Anglo Step||MA Intact||MA Step||Total|
|Agreed to be interviewed||61||47||55||29||192|
|Eligible but refused||4||8||6||11||29|
|Refused, unknown eligibility||0||15||0||12||27|
|Bad phone/unable to reach||5||12||6||43||66|
|Total Families Contacted 1||96||114||138||126||474|
|Eligible Families Contacted 2||72||70||115||52||309|
|Net eligible families contacted 3||65||70||61||52||248|
|Refusal rate 4||5.5%||32.9%||5.2%||44.2%||18.1%|
|Net response rate 5||93.8%||67.1%||90.2%||55.8%||77.4%|
1Total Families Contacted = Number of families contacted by telephone by PAYS staff after initial contact by school recruiter or after returning a postcard indicating interest. Includes not eligible families and families not recruited because cell was full, but excludes those with bad phone numbers and those we were unable to reach.
2 Eligible Families Contacted = Removes confirmed Not Eligible Families from Total Families Contacted (but does not remove refusals with unknown eligibility).
3 Net Eligible Families Contacted = Removes Cell Full Families from Eligible Families Contacted.
4 Refusal Rate = Refusals/Eligible Families Contacted.
5 Net Response Rate = Agreed to be Interviewed/Net Eligible Families Contacted.
Teacher and School Data
During the interview, adolescents provided two teachers’ names for project staff to contact for their participation in the study. At ASU, with the mother’s and adolescent’s written consent, a letter describing the project and requesting participation, a copy of the signed authorization form, $5 compensation, a pre-paid envelope, and a questionnaire regarding the adolescent’s behavior were mailed to each teacher. Follow-up phone calls were made if the questionnaire was not received by the project staff within two weeks. Questionnaires were sent for 200 of the 201 families. The remaining family did not consent to teacher data collection. Of the 400 questionnaires sent, 387 (97%) were completed. Of the 201 adolescents, 197 (98%) had a report from at least one teacher, and 190 (95%) had reports from both teachers.
At UCR, a movie ticket in gratitude for their participation was substituted for the $5 payment to each teacher. A follow-up letter was sent if the questionnaires were not returned after three weeks. Questionnaires were sent for 191 of the 192 families. The remaining family did not consent to teacher data collection. Of the 382 questionnaires sent, 263 (69%) were completed and returned. Of the 192 families, 92 (48%) received two teacher reports, 79 (41%) families received one teacher report, and 21 (11%) families received no teacher reports.
With permission of the entire sample, grades, attendance, test scores and disciplinary actions is being obtained directly from schools, in electronic form from the UCR schools and in manual form the ASU schools.