To study each municipality we used a case study design. We studied individual municipalities, collecting information about each of them through interviews and documents to understand how characteristics and conditions in each municipality attracted and maintained the young adult population. Yin (2003) calls this an explanatory case study. We followed specific historical and structural definitions for our case studies (Stoecker, 1991). The structural boundaries are the geographical boundaries of the municipality with some allowance for its influence by and on the people who live outside of the formal boundaries but interact with the municipality in regular and important ways (such as shopping, working, learning, worshipping in the municipality). The historical boundaries of each case were 1990 to the present.

We also used a comparative case study approach. Ragin (2014) discusses case-oriented and variable-oriented approaches to research, arguing that the comparative case method allows for both. In a comparative case method, the researcher collects information on variables they believe to be important to a single case, while also being open to other information that may be unique to that case. The process can be systematized in a grounded theory approach (Glaser and Strauss, 1967; Charmaz, 2006). Grounded theory simply means that we let the conceptual categories and themes emerge from the research data we gather, rather than trying to rigidly apply a theory to our data. In our case, the theory of why a place may be attractive to young adults is generated by our analysis of participant responses. In grounded theory, the researcher may start with no or a few variables in the first case, listening and watching for other variables that may become relevant. In approaching the subsequent cases, the researcher includes those new variables in the data collection and adds variables as they appear in the data. Grounded theory researchers say that saturation has been reached when they begin to hear the same kinds of responses from participants and, consequently, no new variables are surfacing, which can happen after only a handful of interviews (Guest, Bunce, and Johnson, 2006). When saturation is reached, it is unlikely that new categories or themes will emerge from more participants.

Similar to grounded theory, we started with an initial set of variables, while being prepared for new variables to appear and be added into the analysis. Various rural observers—some with big data and others with deep experience—propose a list of variables that include good career prospects (especially for starting a business), quality local schools (including those that serve non college-bound youth), high speed Internet, peer networking opportunities, support for diversity (including immigrants), programs designed to recruit young professionals, quality outdoor activities, creative class amenities (following Florida, 2002) and civic engagement (Mills and Hazarika, 2001; Ferry 2006; Mcgranahana and Wojana, 2007; Carr and Kefalas, 2009; Center for Rural Affairs, 2012; Schroeder, 2012; Radke, 2013; Petrin et al., 2014). Data for some of these factors were available in pre-existing form, and we collected others from interviews and the core groups (described below).

This is quite a laundry list, however, and simply using a checklist would not be adequate to predict whether a community would attract young adults. Our suspicion was that there may be various combinations of conditions that attract and retain various combinations of young people. Therefore, it was important to not just study whether a place has the resources on the checklist, but to study how young people in those places perceive and interact with those resources.

We knew that gaining access to data about these municipalities would be easier if we were able to build trusting relationships with local community leaders and honor local knowledge. One of the most effective ways of building such relationships and honoring local knowledge is to invite people who would normally be only passive subjects of research to also be co-designers of the research, following the best practices of community-based research (Stoecker, 2013). To do this we attempted to organize a core group in each community that would guide our research and check its accuracy. We began by contacting local UW Extension educators in the county of each identified municipality. We asked them to help bring together people in the municipality who could help operationalize variables, inform data gathering procedures, and suggest people to interview. We also engaged the core groups in reviewing rough drafts of the case study reports in each of their communities to identify and correct potential errors in them, adapting a process called member checking or respondent validation (Mays and Pope, 1995; Buchbinder, 2011). The core groups ranged from 2-6 people in each municipality and often included school officials, local government officials, Chamber of Commerce leaders, and other community members.

Each core group chose from a long list of interview questions approved by our university institutional review board (see Appendix B for the complete list of questions). There was a great deal of overlap in the questions the core groups chose. We then began interviewing individuals and groups in these communities. It is important to understand that we used an open-ended interview method, not a survey method, in collecting data for these case studies. We did this for two reasons. First, with the typical forced-choice survey, the researcher's biases figure heavily in the survey. People will only answer the questions they are asked. If the researcher asks the wrong questions, they will get the wrong answers. In an interview the researcher can ask a question and then have a conversation with the interviewee about what that question means and whether it is the best question to ask (Rubin and Rubin, 1995). The second reason we chose an interview method over a survey is that there is too little past research on why young adults choose to locate in certain less urban places. In a survey you need to know all the possible answers to a question beforehand so you can list them on the survey. Even having community members help construct the survey may not solve this problem as the community members may also not know all the possible answers either.

Our Method

  • Local guidance
  • 12 communities
  • 210 interviews
  • 200+ hours of interviewing

But choosing interviews over a survey involves a trade-off. With interviews we get richer, deeper information, but because of the time it takes we also get information from fewer individuals. We partially compensated for this weakness by a kind of crowd-sourcing method that is gaining popularity in social research. We could have, for example, found only young adults and asked them only about why they in particular chose to live in the community. Instead, we also asked a wide variety of people in the community why they thought young adults in general chose to live in that community. This is similar to research that asks people who they think will win an election rather than who they voted for. Such research is often a more accurate predictor of what is really happening (Graefe, 2014), and it can produce accurate results even with a "biased" sample (Rothschild and Wolfers, 2013).

In total, the research assistants spent more than 200 hours interviewing 210 people across the state.The total number of people interviewed in any municipality ranged from 11 to 32. We generally had more interview participants in larger communities. And, in each case, the researchers believed they reached saturation fairly quickly, so even with what may seem to be a small number of interviews, we are confident in our findings.

We ultimately completed 12 case studies, with at least one in each region of the state. We did not complete case studies in three communities. In one community we were unable to generate enough support for the project to form a core group. In another we were unable to recruit enough interviewees. In the third case we simply lacked the time and resources to complete it, and reasoned that it was likely going to be redundant anyway because it was only a short ways up the road from a completed case study community.