publisher colophon


Where to Go, What to Ask

Selecting and Designing the Case Studies

We are all special cases.

—Albert Camus, The Fall (1956)

We are our choices.

—Jean-Paul Sartre, Being and Nothingness (1953)

Case-study research often walks a fine line between the particular and the abstract. This balancing act arises precisely because the very qualifier case in case study suggests that “the phenomena under investigation... can be found in other places. ...The case may be unique but is not singular” (Castree 2005, 541). The iterative relationship between the concrete and the abstract, between empirical data and theoretical insights, is not only a defining feature of the approach but also the source of its analytical strength. Unlike generalization in the statistical sense, which is largely concerned with representativeness, replication, and external validity, explanations that emerge from case-study analysis involve repeated cycles of crafting theory and concepts that are “rooted in the concrete aspects of the case yet sufficiently abstract [so] that others in similar situations can see how they might apply to their own context” (Baxter 2010, 96).

Because cases are meant to be not isolated events but archetypical enough to offer insights applicable elsewhere, selection is crucial. For us, case selection was complicated because we were interested in two broad and, we hoped, interrelated phenomena. The first was the existence (or not) of diverse and dynamic regional epistemic communities—something difficult to predetermine, since that is the theoretical frame we were applying to processes others might call regional collaboration, regional networks, or regional leadership. There was, in short, no region shouting out that it had built a diverse and dynamic epistemic community, and hence no easily identifiable universe of possible cases to draw from.

The second phenomenon we were interested in was how such epistemic communities might impact social equity and economic growth. Indeed, our curiosity about epistemic communities grew from our interest in understanding whether the structure and dynamics of regional knowledge-sharing networks and the evolution of social norms over time actually shaped patterns of growth and equity—and whether the existence of diverse and dynamic regional epistemic communities could in fact help explain those instances when growth and equity go hand in hand.

In what follows, we explain the specific data we used and the analytical process we undertook to identify what ended up being eleven case studies. As it turns out, it took months for us to settle on the cases—partly because we were going at it from two contrasting directions, and partly because it was so much fun to produce typologies, maps, and data runs (although perhaps not as much fun for the research staff assisting us). We will, however, describe it in a way that’s more succinct than the process—a way that we hope will explain our rationales—and then provide more details about how we conducted each case study.


Why Regions?

Before we dive deep into the details of our case-selection process, it’s important to deal with a preliminary question: Why study regions at all? In chapter 1, we emphasized some of the economic reasons for the emergence of metropolitan regions as an important element of the global economy and also stressed that the regions are one place where the equity argument has been gaining ground (with good reason, as the econometrics of chapter 2 would seem to indicate). Here, however, we want to emphasize the epistemic reasons for looking at metropolitan regions.

Economically, as a wide range of research over the past few decades has shown, regions are a critical scale for processes related to economic growth and change (Acs 2000; Scott 1998; Storper 1997). Processes of innovation, knowledge sharing, and tacit knowledge development; dynamics of growth and decline in economic clusters; efficiencies of daily commuting and goods transportation systems—these are all important processes shaping economic growth that depend on face-to-face communication and the kinds of coordination of actions that only physical proximity allows (Benner and Pastor 2012; Pastor et al. 2000; Porter 1998; Saxenian 1994). Moreover, while causes of social inequality are in large part shaped by regional development patterns and policies, there are promising solutions to these patterns of social inequity that can be achieved through addressing regional political, planning, economic policy, and governance processes; and the political will needed to achieve these solutions can be built at a regional scale, as the ability to have face-to-face communication helps build bridges across the racial and spatial divides that all too often constrain progress (Orfield 2002; Pastor, Benner, and Matsuoka 2011; Powell 1999; Rusk 2001).

There are also important epistemic reasons for our regional lens. In the context of the national economic, inequality, and political crises that we laid out in the first chapter, it is useful to think about regions as living laboratories, with multiple and diverse actors experimenting with how to respond to the dramatic changes we’ve experienced in the past thirty years. Despite (or perhaps because of) substantial variations in initial conditions and clearly different strategies and trajectories, lessons for addressing our national challenges can be gleaned from comparing the relative success or failure of different trajectories. This is particularly so because there is evidence that regional trajectories are diverging more than in the past (Drennan 2005; Scott and Storper 2003; Woo, Ross, and Boston 2015).

Table 3.1 illustrates this increasing heterogeneity among US metros by looking at the coefficient of variation for absolute percent decadal change in a measure of economic health (number of jobs) and a measure of income inequality (the Gini coefficient). The amount of variation among metros fell slightly in the 1990s, but it jumped significantly for change in employment and change in income inequality from the 1980s to the 2000s (the higher the coefficient of variation, the more variation in regional trajectories).


Of course, part of this rising variation could be due to the experience of broader census-designated regions (i.e., Northeast, South, Midwest, and West); that is, maybe it’s not that Atlanta and Denver have diverged but that the South and the West have. Table 3.2 tries to control for metro membership in the larger census region and shows that while in some cases the variation in changes in employment and inequality is lessened when restricted to the census-designated region, there is still a substantial increase in heterogeneity within broad regions of the country. This is particularly noticeable in the Midwest, where there was far more variation among changes in income inequality in the 2000s than in the 1980s and 1990s. We see a similar trend of variation in changing employment trajectories in the West. All of this is to say the following: it is clear that individual metropolitan areas are experiencing economic changes in ways that do not necessarily coincide with national or larger regional trends; that these differences are increasing over time; and there may be lessons to learn from how each region is regrouping and responding.


Choosing Regions

While all the analysis above suggests that regions are an important scale for learning, we are still left with the task of selecting cases to maximize theoretical insights and most effectively generate generalizable principles. If we could have selected regions based on the independent variable—the strength, diversity, and dynamism of regional epistemic communities—we would have used this as our criterion and investigated in those regions how these knowledge communities shape the dynamics of growth and equity. Without being able to easily identify such cases, however—and since our goal was partly to develop greater sophistication in our understanding of the very concept of diverse and dynamic epistemic communities—our case-selection process is best understood as a theoretical sampling approach, that is, an iterative process of selection with the goal of gaining a deeper, richer understanding of a concept or theory across a range of different contexts and conditions (Glaser and Strauss 1967).

Thus, we ultimately aimed to select three kinds of cases: first, regions where equity and growth came together, and where we could then investigate the nature of epistemic communities in the region and gauge what influence, if any, they might have had on those regional trajectories; second, regions with substantially below-average growth and social equity metrics, to see whether we could find evidence of a lack of diversity and dynamism in a regional epistemic community (or evidence of a thriving epistemic community despite poor concrete results, a clear counterexample to our framework); and third, regions that had reputations (from previous research by ourselves and others) of having strong, collaborative regional processes that did not necessarily show strong evidence of equity and growth going together, a pattern that might challenge our presuppositions about the positive impacts of such collaborations.

Given these various criteria, our first step in selecting case-study regions involved selecting some high performers, with the goal of eventually visiting those regions and investigating the extent to which the good results were due to structural factors, dumb luck, or the role of knowledge communities. In some ways, this was a replication of the approach taken in Just Growth, in which we moved strictly from data to theory. There were, however, a few modifications introduced in this exercise. The first change was that we had previously looked at patterns only in the 1980s and 1990s; for this study, we updated our analysis to include the 2000s. A second modification, which we explain in more detail below, is that we considered here the end-point as well as the trajectory; that is, we accounted for the eventual level and distribution of income as well as the shift over time in key related variables. In this way we could target regions that were truly high achievers by the end of this thirty-year period, not just regions that might have improved, albeit only slightly, over time.

Our universe of regions for consideration consisted of the US metropolitan regions that had at least 200,000 residents in 2000. Using this break, we ended up with 192 regions in our sample. To identify broad patterns of economic growth and social equity, we examined patterns over four time periods: the 1980s, the 1990s, the 2000s, and the entire thirty-year period (1980–2010). Following the methodology we developed in Just Growth, the indicators we used to measure economic growth were the change in employment and the change in earnings per job—because we wanted to identify not just growth in jobs but growth in good jobs. The indicators we used to measure equity were the change in the income gap and the change in the percentage of the population living below poverty—because it does no good to close the income gap by making everyone poorer.

To measure the income gap, or income inequality, we used the 80/20 household income ratio, which compares the 80th percentile of income earners with the 20th percentile, with higher ratios indicating more inequality. Recent attention has focused more on the tails of the income distribution, particularly the top 1 percent (Alvaredo et al. 2013). We focused on the 80/20 ratio, partly because we wanted to look at broader measures of social distance but also because calculations of a different and wider ratio (for example, the 90/10 ratio) were less reliable with the 1980 census data.1

While we were interested in looking at the changing dynamics over time, we also wanted to provide a snapshot of the metro regions’ economic and social well-being at the end of the thirty-year period, in 2010. That is, where did they end up when all was said and done? After all, improvement from rock bottom might be impressive, but if the region remains poor and highly unequal, that’s, well, a bit less impressive. To measure final economic well-being we used the median household income, and to measure equity we used the Gini coefficient for household income, a standard indicator of income inequality. These variables had the virtue of being both easily collectible and not the same variables used in the trajectory analysis, thus providing a more independent measure of the end state.

For our longitudinal analysis, we drew the economic growth variables (change in employment and earnings) from the US Bureau of Economic Analysis Personal Income Accounts Tables, and the equity variables from the American Community Survey. For our 2010 analysis—the snapshot of the economic and social well-being of metros—we used data from the American Community Survey, which we downloaded from the Missouri Census Data Center.

Making a First Cut

In order to more easily compare regional performance across these four indicators and four different time periods, we created two separate composite indices: a growth index—combining the percentage change in both employment and earnings per job—and an equity index—combining the percentage change in both poverty and the 80/20 household income ratio. While these indices were used to compare all metros, there are broad regional differences across the country, so we benchmarked all of the metros against their respective larger US census-designated region: Northeast, South, Midwest, or West. To do this, we normalized each measure into detrended z-scores: we grouped the metros by census-designated region, and for each record, we subtracted the census regional mean and divided by the census regional standard deviation.

From this process, we obtained sixteen z-scores for each region: one for each of the four variables in each of the four time periods (1980s, 1990s, 2000s, and the whole thirty-year period).2 We then computed the growth index as the mean of the eight growth-related z-scores, and the equity index as the mean of the eight equity-related z-scores (appendix A offers an account of those indices across all 192 regions, as well as a look at some of the initial and ending variables that go into the calculations and case selection). Finally, we ranked both the growth and equity indices into terciles (best, middle, and worst) across the entire 192-region sample, and mapped the results to observe the distribution of regions based on their relative scores on these indices (map 3.1). Together, the two tercile scores form a nine-cell grid, and the upper-right-hand cell of that grid is the “best,” in that it indicates both solid growth and relatively good trends in equity (meaning either actual upticks, or given the general trends in the United States during this time period, better than whatever dismal average was achieved by one’s particular census region).

MAP 3.1. US Metro Regions by Growth and Equity Indices, 1980–2010s.

As we mentioned above, we also wanted to go beyond trends over this thirty-year period and investigate whether the ending point on income levels and distributional measures was also above average. After all, if a region started off from a particularly low point in 1980, it could appear to improve dramatically over that period but still end up substantially below average—that is, the relative acceleration in its improvement may simply reflect a process of reverting to the mean (and maybe not even getting there). So, to make sure we were capturing regions that not only posted above-average improvements over time but also ended up at above-average levels of income and equity in the end, we looked at the median household income and the Gini coefficient of household income.3 In keeping with our trajectory analyses, we also normalized these endpoint scores by the broad census region, producing two z-scores (one for equity and one for growth) which could then be ranked into terciles. We show the results in map 3.2.

Those regions that are truly the strongest-performing in terms of both change over time and end result are in the upper-right corner of the distribution in both maps, 3.1 and 3.2. The number of regions that fall into both of those categories is interesting, but also relatively small. It includes no metro regions in the West; only Manchester-Nashua and New Hampshire in the Northeast; four relatively small and non-diverse regions in the Midwest (Des Moines and Cedar Rapids, Iowa, and Madison and Appleton, Wisconsin); and three regions in the South (San Antonio and Killeen–Temple–Fort Hood, Texas, and Raleigh-Cary, North Carolina). While we considered all of these regions for possible case studies (and ended up choosing two of them for our final list), we were concerned that they represented a relatively narrow experience of metropolitan regions in the United States.

MAP 3.2. US Metro Regions: Relative Levels of Household Income and Gini Coefficient of Household Income in 2010.

Narrowing Down

Given the desire to expand the range for the case studies with positive growth and equity metrics—and knowing that it was extraordinarily difficult to actually maintain such high performance on dual criteria over multiple time periods—we decided to loosen our cutoffs for consideration on these metrics. We did this in two ways. The first was to take as successful those regions that fell into the top half—rather than the top third—in each of our growth and equity measures for both the change over time and end point. The second was to consider regions that excelled (were in the top third) in either growth or equity and were average or better on the other metric. To understand this visually, we are talking about regions that are in one of the three boxes on the upper right of the three-by-three matrix in maps 3.1 and 3.2.

The second approach produced a broader set of metros, even when we required that the regional scores wind up in these top three squares for both our change-over-time metrics and our end-of-period metrics. We show the results of that approach in map 3.3.

MAP 3.3 U.S. Metro Regions Scoring in Top Three Squares in Both Over-time and Endpoint Measure of Growth and Equity

We drew a few conclusions about possible case studies from this analysis. But first, it was comforting to know that several of the “just growth” regions studied in our earlier volume stood the test of (more) time, including Jacksonville, Kansas City, and Nashville. Another of the regions we explored in Just Growth, Columbus, does not make the cut in the broad map, but did when we narrowed down to a comparison with just the Rust Belt states of Pennsylvania, Ohio, Michigan, Illinois, Indiana, and Wisconsin. This suggested to us that both the method and the underlying theory might be robust to different periodization—and with that, we went more confidently ahead in this phase of the selection process.

Looking at the full set of maps, one immediately sees two regions as obvious candidates for study: San Antonio, Texas, and Raleigh-Cary, North Carolina. Both regions score in the top tercile in all of our metrics while also being large and ethnically diverse enough to provide potentially interesting lessons for a range of other metropolitan regions around the country. Another region that emerged as a candidate was Salt Lake City, Utah. While it is the neighboring region of Provo-Orem, Utah, that actually shows up most prominently in these maps, Salt Lake City does show up in map 3.3, and locals there tend to think of the region as the entire Wasatch Front, including Provo, Salt Lake City, and Ogden-Clearfield to the north. Since Salt Lake City is the urban core that drives change in this broader region, it made sense to focus on it. Finally, Seattle showed up as a top performer in terms of both trajectory and endpoint in map 3.3—so it seemed an obvious choice.

We were also interested in selecting a region within the broad Rust Belt. Once serving as the industrial heartland of the country, the area from western New York and Pennsylvania through eastern Wisconsin and Illinois has suffered some of the worst effects of the decline of manufacturing and broader economic restructuring of the past thirty years. We thought that examining a region that had weathered this transition relatively well could provide interesting insights. In looking at the patterns of change over the thirty-year period, as well as the status of income and inequality in 2010, Grand Rapids, Michigan, stood out. It was the only metropolitan region in the entire Rust Belt to score in the top three squares in our three-by-three matrix for both change over time and status in 2010 (see map 3.3), so we selected it.

As we explored these patterns, we also realized that it would be interesting theoretically to identify cases that were not just solid performers across the entire thirty-year period but had also experienced significant shifts in growth and equity during the same period. If a region were able to dramatically reverse a period of economic decline and growing inequality, and jump onto a path of faster economic growth and improving equity, it could provide important insights into the factors that shape the regional ability to recover from economic shocks. To investigate this, we created typologies describing the trends that occurred in different metros between 1980 and 2010, classifying each of the four measurements of change (change in employment, change in earnings, change in the income gap, and change in the percentage below poverty) as either “good” or “bad” in each of the four periods (1980s, 1990s, 2000s, and the thirty-year period).

Specifically, we labeled each growth measure “good” if it was above the median value of that measure in its respective census-designated region, and “bad” if it was below the median. Conversely, we labeled each equity measure “good” if it was below the median of that variable in the respective census region (i.e. relatively less inequality and poverty), and “bad” if it was above the median (i.e. relatively more inequality and poverty). Using these tags, we identified whether a metro had improved or declined in terms of growth and equity in each of the four periods, and we created a typology to categorize the metros:4

•  consistently good: good in 9 of the 12 changes that occurred in the 1980s, 1990s, and 2000s (with the condition that at least three-quarters of the changes in two of the decades were good and at least one growth measure and one equity measure was good in the remaining decade)

•  thirty-year good: good in all four changes in the thirty-year period

•  bounce back: bad in at least three-quarters of the changes in the 1980s and 1990s and good in at least three-quarters of the changes in the 2000s, or bad in at least three-quarters of the changes in the 1980s and good in at least three-quarters of the changes in the 1990s and 2000s

•  thirty-year bad: bad in all four changes in the thirty-year period

•  consistently bad: bad in 9 of the 12 changes that occurred in the 1980s, 1990s, and 2000s (with the condition that at least three-quarters of the changes in two of the decades were bad and at least one growth measure and one equity measure was bad in the remaining decade)

Within this typology, the regions that had been able to recover from an economic shock and achieve positive growth and equity dynamics in subsequent periods are those in the bounce back category. By selecting cases of this type, we hoped to explore whether the observed bounce back was simply a function of structural factors or whether more conscious forms of knowledge generation and cross-sector collaboration—that is, the formation of an epistemic community—had played a role. Among the twenty or so candidates in this category, Oklahoma City and San Antonio were particularly interesting. San Antonio was able to turn around in enough time (going from bad in the 1980s to more promising in the 1990s and 2000s) to land in our top tier in 2010 in terms of growth and equity trajectories (map 3.3). Oklahoma City remained at the bottom, but it had moved from having bad outcomes in the 1980s and 1990s to much better ones in the 2000s, leading to curiosity about the turnaround.

As we examined these different categories in more depth, another striking feature stood out. There are very few states in the United States that had metropolitan regions that fell into both the best (consistently good or thirty-year good) and worst (consistently bad or thirty-year bad) categories. This may be partly a function of the state size needed for variation: California and Texas, the nation’s two largest states, are among the three states that have regions that fell both into the worst and best categories of change over this thirty-year period. The other, however, is North Carolina, which is especially intriguing since it just makes the top ten states in terms of population. In any case, the general lack of extremes within states made us wonder about the role of state policy and whether it might be useful to control for similar state environments.

Considering this, we decided to develop a set of paired comparisons across two different states. Focusing on North Carolina and California, we selected regions that fell into two different categories: regions with strong patterns of both economic growth and improved social equity across the full thirty-year period (Sacramento in California—despite its dip in performance in the 1990s—and Raleigh in North Carolina); and regions that were among the worst performers in growth and equity in each state (Fresno in California and Greensboro in North Carolina). We wound up adding an additional case in each state, for reasons discussed below, so we actually had state triads, a sampling strategy that provides even more information.

Theoretically, we were also interested in identifying regions that had reputations for collaboration that might be an indicator of a diverse and dynamic epistemic community, but in which positive patterns for growth and social equity metrics failed to emerge (or were mixed). In looking at our two-state, two-region comparison, we realized that there were prime opportunities for selecting a third region in each state that met this criterion. In California, Silicon Valley’s economic success has been linked to both its open, flexible, and dynamic labor markets and its particular regional culture of collaboration (Saxenian 1994). In North Carolina, Charlotte had a strong reputation in the 1980s and 1990s for regional collaboration and relatively progressive (by Southern standards) social norms, all part of repositioning the region as a central logistics and banking center for the country and an example of the New South (Pastor et al. 2000). Thus, we decided to add Silicon Valley and Charlotte as cases. Reputation for strong regional collaboration also helped reinforce our reasons for selecting Salt Lake and Sacramento, both of which were well known for regional planning processes.

Having finished this range of considerations, we wound up with a total of eleven cases—Charlotte, Fresno, Grand Rapids, Greensboro, Oklahoma City, Raleigh, Sacramento, Salt Lake City, San Antonio, San Jose/Silicon Valley, and Seattle—which fell into a series of different categories. Six regions (Grand Rapids, Raleigh, Sacramento, Salt Lake City, San Antonio, and Seattle) were included for having positive growth and equity trajectories over the full thirty-year period of our analysis, and we added a seventh, Oklahoma City, because it was an interesting example of a “bounce back” in the latter years; here, we wanted to see whether regional culture, social norms, and epistemic communities played a role in resilience and recovery. Two regions (Fresno and Greensboro) were included because of being poor performers on both growth and equity, but being in states where other regions have thrived; here, we wanted to see whether fractious cultures played a role. Finally, two regions (Silicon Valley and Charlotte) were included because of having evidence of strong regional collaborations but growth and equity trajectories that were less positive than in two other cases (Sacramento and Salt Lake City) that also had well-known regional collaborations.


Once we had selected the regions to study, we began to assemble a broad range of data about them, as well as a list of possible people and organizations to interview in each region (data profiles for each of the cases are offered in appendix B, and a list of the eventual interviewees is offered in appendix C). Because we were interested in not just the outcomes but also the process of community building and knowledge creation, we focused on the most prominent organizations in four types of key constituencies: the private sector, particularly the largest chamber of commerce in each region and other major business associations; the public sector, particularly the regional planning organization and other prominent public-sector officials; labor organizations, particularly the regional labor council and prominent individual union leaders, where they existed; and nonprofit organizations, including community-based organizations and major philanthropic entities.

We then scheduled multi-day visits to all the case-study regions, interviewing as many people in each of our four broad sectors as we could fit into our (and their) schedules. In some regions, respondents were not available in person, but made themselves available for phone calls either before or after the visit. For both the in-person and the phone-call interviews, we developed a detailed interview protocol that guided our discussions. The protocol was designed to help us better understand the nature of regional knowledge networks, including mechanisms of knowledge generation, social norms about collaboration and conflict within regional governance processes, underlying factors that either contributed to collaboration or mediated conflict, and whether and how interactions between actors were extended over long periods of time.

The specific number of interviews conducted in each region varied from 11 to 19, with a total of 172 interviewed and each interview lasting between one and two hours; again, a full list of interviewees in each region is provided in appendix C. These interviews were supplemented with detailed secondary research on regional dynamics from both web-based sources and the academic literature. This additional research occurred both before the visits—to help prepare for our interviews—and afterward, since the in-person interviews inevitably raised new questions or illustrated new dynamics worth investigating in more depth.

Cultures of Conflict and Collaboration

While there are finer nuances of regional culture and knowledge sharing that we were not able to capture in the time available for each case study, our goal was not to write comprehensive historical studies of any single region. Rather, our goal was to investigate broad patterns of regional governance—particularly to understand the ways in which diverse constituencies are involved in or marginalized from regional governance processes—and how conflict and collaboration have emerged and evolved over time.

The conflict piece was particularly important since we realized that any regional process had to effectively address divergent interests; we were particularly struck by the admonition of Lester and Reckhow (2013) to understand that progress on equity was often driven by “skirmishes.” To get at this, specific questions were focused on:

•  sustaining cultures of cooperation, in which we asked respondents to identify the long-term historical processes, social factors, and interpersonal social norms that have helped shape or shred cultures of collaboration within their region

•  learning to collaborate, in which we asked respondents to describe, against that broader backdrop, an example of successful collaboration across diverse constituencies in the region, and to share their perspectives on what contributed to the success of that collaboration

•  dealing with conflict, in which we asked people to describe an instance of major conflict in the region between different constituencies, and to share their perspectives on what the conflicting goals and values were that shaped that conflict, how it was or was not resolved, and what lasting impact that had on regional governance processes

•  sources of knowledge, in which we asked respondents how they obtained information and knowledge about, first, economic trends and, second, social conditions in the region, with an eye to seeing whether there was, in fact, a shared knowledge base across constituencies.

In all of the interviews in a region, we were looking for commonalities, as well as inconsistencies, in how people portrayed processes of collaboration and conflict in regional governance processes. What are the processes that help people develop a common language and cognitive frames that allow them to communicate effectively and share knowledge across diverse experiences, values, and priorities? When do such processes emerge, and how? What factors preclude those processes from developing? And who holds influence and power in regional governance processes?

Some Additional Preliminaries

In the following chapters, we present the insights that emerged from each of our case studies. Before jumping into that analysis, however, it is worth highlighting a few cross-cutting methodological issues that shaped our analysis. First, while we used the official US Census definitions of metropolitan statistical areas (called core-based statistical areas, or CBSAs) for our data analysis and case-study selection, in many of our cases respondents defined their functional region as a somewhat different geography.

In Sacramento, for example, the official CBSA is a four-county region (El Dorado, Sacramento, Placer, and Yolo Counties), but the regional council of governments covers two additional counties (Yuba and Sutter); this six-county area is the one most frequently cited in discussions of regional collaboration. Similarly, in Salt Lake City, the official CBSA consists of Salt Lake County and the sparsely populated Tooele and Summit Counties to the west and east of Salt Lake County, but most of our respondents talked more about regional integration north into neighboring Davis and Weber Counties (part of the Ogden-Clearfield Metropolitan Statistical Area) and south into Utah County (part of the Provo-Orem Metropolitan Statistical Area).

While actor perceptions of the relevant regional borders could sometimes differ from the official CBSA boundaries, in nearly all cases the patterns of growth and equity dynamics in these neighboring areas were similar enough to the data for the official CBSA that we were not concerned about gaps between our characterization of regional dynamics and the categorizations that might be offered by regional respondents. The one exception to this was in the Raleigh-Cary metro. Here, residents of the region consider the functional region to be Raleigh-Durham, but in official census designations there are two distinct CBSAs. Raleigh is a three-county metro covering Franklin, Johnston, and Wake Counties, and exhibits the strong growth and equity results that drove us to select the region, while the neighboring Durham–Chapel Hill metro covers four counties (Chatham, Durham, Orange, and Person), includes the largest concentration of African Americans in the region, and has much worse patterns of social equity than Raleigh (although its growth performance was strong).

Did this mean that Raleigh was simply shifting its problems over to neighboring Durham? To make sure our identification of Raleigh as a strong performer wasn’t simply an artifact of a Census definition of the region that didn’t correspond to local understandings of the functional region, we recalculated the equity and growth indices, as well as the median household income and Gini coefficient in 2010, for the combined seven-county area. As it turned out, the newly combined region still remains in the top third of both equity and growth indicators. In the statistical table we provide in appendix B about the Raleigh case study, the data is for this combined seven-county area, while for each of the other case studies, we present data for the official Census CBSA definition of the metro.

Second, as described above, our case-study selection was in part driven by consideration of the role of the state context in shaping regional dynamics. This is most obvious in our two-state, three-region paired comparison cases, but implicit in the other cases was an understanding that state policy might shape regional dynamics. In our interviews, however, it was clear that state policy had very little to do with people’s understandings of regional dynamics, so we end up saying very little about it in the specific case-study descriptions. Where state policy was relevant (as in North Carolina, where new state laws have eroded the ability of local municipalities to annex and so restrain suburban fragmentation), we raise this. Overall, however, the patterns of regional development in each of our cases were primarily understood by our respondents as being driven by processes and relationships specific to the region, and this is our focus in what follows.


This chapter has described the way we went about selecting regions to help us understand the relationship between epistemic communities, regional social norms, and patterns of economic growth and social equity. We essentially approached this task from two directions. On the one hand, we started from the patterns of growth and equity, identifying regions with a range of patterns (overall good, overall bad, bounce-back) and choosing a range that could help us explore the role epistemic communities might play in explaining those patterns. On the other hand, we identified regions with patterns of collaboration that might be indicative of diverse and dynamic epistemic communities but with patterns of growth and equity that diverged, including from our hypothesis that such epistemic communities should contribute to both strong growth and improved social equity.

Since our theoretical interests in the book are primarily about the nature and impact of these regional epistemic communities, we decided to group them in the chapters that follow not so much on outcomes as on processes. Thus, in the chapters that follow, the regions are clustered into the following four categories.

1. Planning-influenced community building. Here we include Sacramento and Salt Lake City, both places where there were very conscious efforts to bring constituencies together around regional growth and improvements (the Blueprint process for the former and Envision Utah for the latter). These cases provide insight into how formal processes can create and sustain diverse epistemic communities over time, but also how such processes—particularly in Sacramento—can leave some issues of equity to one side, but then exhibit the dynamism to be more inclusive.

2. Elite-driven regional stewardship. Here we include Charlotte, Grand Rapids, and Oklahoma City. As we will see, the first two cases share the characteristic that their good performance has faded, suggesting that one needs to go beyond elites and top-down approaches to be more effective and inclusive. On the other hand, the turnaround in Oklahoma City is remarkable, and there is significant evidence of attempts to widen the circle of participants in and beneficiaries from metropolitan development.

3. Conflict-informed collaboration. Here we examine Fresno, Greensboro, and San Antonio. It is only the latter where we actually observe conflict feeding into what later becomes well known as a culture of collaboration; in the other cases, we essentially see a continuing war of attrition between competing and distant social actors.

4. Knowledge economies and networks. Here we include Raleigh, Seattle, and Silicon Valley. We suggest that Silicon Valley was getting it right—but is now getting it wrong, as a more rootless group of entrepreneurs eschew the practices of what were once the region’s stewards. Raleigh has woven together a seemingly coherent epistemic community. Virtually every key leader repeats the same mantra about the region’s Triple Helix growth framework. Meanwhile, Seattle has made a remarkable commitment to maintaining equitable opportunity even as it is—like the others—subject to the highly disequalizing trends associated with being a center of innovation for the “new economy.”

Before diving in, we want to re-emphasize one key aspect of the analysis that follows. In all of our case studies, the focus was on processes of governance, not formal government policies or the creation of new decision-making institutions. In some cases, as the literature on epistemic communities would suggest, processes of repeated interaction and knowledge sharing on a regional basis ended up being institutionalized through specific organizations that facilitate the development of regional epistemic communities. We find this in our examples of planning-influenced community building: the evolution of the Sacramento Area Council of Governments as a provider of the infrastructure for regional knowledge-sharing networks, and the development of Envision Utah, a small nonprofit organization that plays a similar role in Salt Lake City.

But even in these cases, there is no single institution that can be considered the true core of the regional epistemic community. In all cases, rather, there is a diversity of organizations that help facilitate more informal information sharing and knowledge development across many different constituencies. In places where this is most widely developed, there exists a philosophy of diverse knowledge-sharing across acknowledged partners, rather than a single organization that institutionalizes the diverse regional epistemic community. Thus, for example, in Grand Rapids, multiple respondents talked about the “four-legged stool,” in which four distinct organizations had a well-understood division of labor in shaping regional developments, with strong cross-organization communication and collaboration.

Similarly, in Raleigh-Durham, there is no single organization underpinning the Triple Helix that regional residents identify as critical for their success, but it is exactly that frame that gets repeated over and over. In Seattle, government action is important, but the usual reference by interviewees was to the multi-stakeholder negotiations known as the Seattle process, which civic leaders go back to again and again as a way to resolve conflict. In San Antonio, the celebration of collaboration between sectors is so much a part of the region’s self-presentation that it is right to worry whether the important role of political skirmishes and social movements in putting equity on the agenda will be forgotten. In short, while policies may be implemented and institutions created, the real underpinnings we found (and were searching for) have to do with the evolution of regional social norms about knowledge generation, information sharing, and conflict resolution.

None of this is as easy to measure as, say, a Gini coefficient. We are definitely aficionados of large data sets, complex metrics, and multivariate regressions (remember chapter 2?), and we did want a quantitative approach to both inform and set the stage for our case studies. But there is a sort of depth of understanding, particularly of qualitative social processes, that can only be attained with visits to the field. And so we went, with questionnaires in tow and frequent-flyer accounts in hand, as we crisscrossed the country in search of community. What we learned sometimes confirmed and sometimes surprised, but it always informed, as we looked for new ways in which leaders were sinking roots, forging relationships, and bringing reason to a conversation about our metropolitan future.

3.  Where to Go, What to Ask: Selecting and Designing the Case Studies

Choice and Consequence

The Quest and the Questions

Ready to Launch

1. The reason is that we built up the various income ratios for the metro regions by adding up data from the census tracts. Because we worked with summary data, not micro-data, this data is actually recorded as the number of households in various preset income categories. To get at the exact income breaks for the 80/20 ratio at a regional level, we summed the number of households in each break across tracts for each metro area, approximating the distribution of household income. We then identified the income brackets containing the 80th and 20th income percentiles and applied a Pareto interpolation to choose the exact level of the 80th and 20th percentiles. This process helped us estimate the tails, but more effectively when there are many brackets; in 1980, the income categories were few in number, so we thought that estimates at the 90th and 10th percentiles would be less reliable than estimates at the 80th and 20th percentiles. The 80/20 ratio is also used in econometric work by Hill et al. (2012), and we took a similar approach in Benner and Pastor (2012).

2. We realize that including the whole time period as well as the various separate decades might seem like double-counting, but essentially we were trying to reward a region for highly consistent performance; see the discussion of the method in Benner and Pastor (2012).

3. While looking at 2010 made sense because it correlated with the time period being used for the trajectory analysis, we were concerned that this might be viewed as an odd and perhaps biased or unrepresentative ending point because of the ways in which the impacts of the Great Recession may have varied across different metropolitan economies. We therefore experimented with several other end points, including 2007 (right before the onset of the Great Recession) and an average of the years 2006–2010 (to smooth out business-cycle effects and perhaps get at a “true” value). As it turns out, no important differences in the characterizations of income and inequality resulted, so we stuck with the more intuitively obvious (and seemingly consistent) 2010. Also, while we considered both per capita income and median household income for our measure of end-point economic well-being, we decided to use only median household income. In part, this was because household income seems to be a better indicator of people’s access to economic resources. But this decision was also made in part because Latino households have a significantly higher-than-average household size compared with other racial groups, meaning that the per capita income figures shifted somewhat disproportionately in regions with a growing percentage of Latino households (a compositional effect). We could see no logical reason to introduce what seemed to us a bias against selecting Latinizing regions, particularly given that the demographic changes going on in the country make the examination of such regions especially important for lessons for the future. Our equity measure was an easier choice. Since we were already using income ratios in our measures of change over time, we decided to use the Gini coefficient of household income.

4. This process resulted in a number of other region types that we ended up not using in our selection process but which might be of interest in other research. These included valley, or up-down-up: “good” in at least three-quarters of changes in the 1980s, “bad” in at least three-quarters of changes in the 1990s, and “good” in at least three-quarters of changes in the 2000s; mountain, or down-up-down: “bad” in at least three-quarters of changes in the 1980s, “good” in at least three-quarters of changes in the 1990s, and “bad” in at least three-quarters of changes in the 2000s; and fall back: “good” in at least three-quarters of the changes in the 1980s and 1990s and “bad” in at least three-quarters of the changes in the 2000s, or “good” in at least three-quarters of the changes in the 1980s and “bad” in at least three-quarters of the changes in the 1990s and 2000s.

Additional Information

Related ISBN
MARC Record
Launched on MUSE
Open Access
Back To Top

This website uses cookies to ensure you get the best experience on our website. Without cookies your experience may not be seamless.