A. Detailed Example of Interactive Approach Used in Service-User Typology Research
McGroder et al. (2003) sought to develop a typology of barriers to employment in a sample of 916 welfare-receiving mothers subject to mandatory participation in a welfare-to-work program. Using data from the National Evaluation of Welfare-to-Work Strategies, the researchers examined a number of variables reflecting mothers' service needs and challenges:
· Mothers' human capital (literacy and whether they graduated from high school)
· Mothers' labor force experience (past employment and length of time receiving welfare)
· Mothers' physical and psychological well-being (health problems, depressive symptoms, and their sense of control over their lives)
· Logistical barriers to employment (access to child care and transportation)
Applying cluster analysis to these nine variables yielded five clusters:
· The largest cluster, "Lower Risk," which comprised 30 percent of the sample, was characterized by below-average scores on each of the nine hypothesized barriers/risks. The mothers in this cluster were the least likely to report transportation or child care problems, had the fewest depressive symptoms, and averaged only one or two years of receiving welfare. None reported health problems keeping them from school or work, most had a high school degree and higher literacy scores than the sample average, and every mother in this cluster reported having had worked full-time for the same employer for at least six months.
· The second cluster, "Lower Risk but No Work History," comprised 16 percent of the sample and scored below the sample average on all barriers to employment except one. Unlike the lower risk cluster, every mother in this group reported never having worked full time for six months or more for the same employer. It was also the factor that distinguished this group from each of the others.
· The third cluster, "Human Capital but Psychological and Logistical Barriers," comprised 19 percent of the sample, and the members of this group were at risk in some ways, but at lower risk in others. As in the first two clusters, most mothers in this group had a high school degree and literacy scores above the sample average, and none reported a health problem interfering with work. Almost all had work experience and received welfare for less than two years. However, this group scored the highest of any group on depressive symptoms and lower perceived control over their lives, and most reported not being able to go to school or work because they could not afford child care.
· The fourth cluster, "Health Barriers," comprised 18 percent of the sample and had only one barrier to employment. Though they scored at the sample average on human capital and employment history, every mother in this cluster reported not being able to go to school or work because they or a family member had a health or emotional problem.
· The final cluster, "Multiple Barriers," comprising 17 percent of the sample, had the poorest human capital, spent the most time on welfare, and over half had never worked full time for six months or more for the same employer. These mothers also had relatively many depressive symptoms and little sense of control over their lives, and most reported problems securing child care and transportation.
Examining maternal employment and child outcomes in each cluster, the researchers found that two subgroups, Multiple Barriers and Health Barriers, had the worst employment outcomes. This suggests that the severity of a single risk factor-family health problems-was as consequential for maternal employment as numerous risk factors for mothers' employment. However, children's outcomes were compromised only in the multiple barriers subgroup; children in the health problems subgroup typically did not look different from those in the lower risk group. The authors conclude that it may not be enough to know which barriers predict longer welfare stays or less employment-rather, it is important to know how barriers naturally co-occur in a given sample in order to target families with the appropriate package of services.
B. Detailed Example of Interactive Approach Used in Audience Segmentation Research
Slater and Flora (1991) sought to segment a sample of 1,669 adult residents from four central California cities into relatively homogenous subgroups reflecting their health orientation in an effort to identify appropriate health messages and intervention strategies for each segment. The researchers analyzed a number of individual and social influences on health behavior:
· Knowledge, including the link between weight and cardiovascular disease, and awareness of the preventability of cardiovascular disease
· Beliefs, including confidence in being able to maintain a healthy diet, confidence in being able to undertake exercise, and beliefs about the cost and palatability of healthy foods
· Social influences, including family and peer norms and conversations relating to diet and exercise
· Health behaviors, including dietary habits, walking, more strenuous exercise, smoking, and drinking
· Demographic characteristics, including age, gender, education, income, marital status, and household size
Applying cluster analysis to these variables yielded four clusters:
· The largest cluster, "Healthful Adults," comprised 39 percent of the sample and was characterized by healthful scores on several dimensions. They were aware of the health risks of being overweight and of the preventability of cardiovascular disease but were not overly worried about their health, were open to healthier habits, engaged in walking and healthful eating, refrained from smoking, were exposed to peers with healthy habits, and had household support for healthful living.
· The next largest cluster, "Unhealthful Adults," comprised 32 percent of the sample and was characterized by mostly unhealthful scores on a few critical dimensions. Members of this group understood the health risks for cardiovascular disease but nevertheless had poor diets, smoked, drank, had unhealthful peers, did not want to change their health habits, and did not feel confident they could change their diet.
· "Worried Older Adults" comprised 24 percent of the sample and had both positive and negative factors influencing their behavior. On the one hand, they understood the health risks for cardiovascular disease, they cared about preventing this outcome, they worried about their health, and they refrained from drinking. On the other hand, they did not believe cardiovascular disease was preventable, did not feel confident they could maintain regular exercise, had unhealthy diets and exercise habits, were daunted by their beliefs that healthy foods were expensive and not tasty, and received little to no household support for a more healthful lifestyle.
· The smallest cluster, "Healthful Talkers," comprising about six percent of the sample, was in some ways similar to the healthful adults, but this group was characterized by their engagement in intense exercise and their tendency to talk about health issues.
Based on the findings of this cluster-analysis, the researchers discussed implications for designing and targeting cardiovascular health interventions. For example, they suggest that interventions targeted to Worried Older Adults should focus on educating them on ways to reduce their cardiovascular risks and increasing their sense of efficacy with respect to exercising-perhaps by providing opportunities to exercise to slowly gain confidence. These adults might also benefit from having household members involved in the program so they could better support the health habits of the Worried Older Adults. Unhealthy Adults, by contrast, may need a more comprehensive set of strategies, targeting not only their unhealthful knowledge, beliefs, and behaviors, but also the norms and behaviors of the their family and peer networks. Finally, Slater and Flora (1991) point out that the Healthful Adults may need no intervention, and yet, their positive health orientation may make them the most likely segment to sign up for (but least likely to benefit from) a cardiovascular health intervention. Moreover, because Healthful Adults comprise the largest segment of the target audience, any program evaluation including this positively selected group is likely to show negligible impacts for the sample as whole, even if such an intervention had positive impacts on one of the other subgroups.