Measurement and Storytelling in Regional Science

Tales from the West Virginia Hollows

The above phrase was in fact the subtitle of Stephan's doctoral dissertation. Stephan first met Andy almost exactly 20 years ago, while he was an Economics graduate student at Berkeley. Stephan had come back from northern Africa for his PhD precisely to study rural economic development in the United States, as his time in the hinterlands of the Atlas Mountains and the Sahara made him realize just how little he knew about the rural economies of his own country. Somewhat to his surprise, he returned to find the topic similarly at the margin of the economics field. Annalee Saxenian was in her early years as a junior faculty member at Berkeley; she provided helpful initial orientation during discussions with Stephan, most especially when Saxenian suggested he chat with a visiting scholar in her planning department, Andy Isserman.

By that point, Stephan had become a bit disillusioned with his rural-labor-market niche, but Andy quickly revived his excitement. Andy's sketch of the Appalachian situation, in general, and West Virginia’s, in particular, helped spur the same intellectual curiosity and social challenge that must have moved initial development efforts in the 1960s. Soon after Andy returned to West Virginia University (WVU), he invited Stephan to join him at the Regional Research Institute for six months as a “visiting scholar” to collect data and, more importantly, get a feel for the labor markets Stephan wanted to model. Stephan leapt at the opportunity.

Those months helped form Stephan as a scholar. Starting with a review of the national and regional literature, he started to make and extend contacts in the state, which led to (at the time) a novel data set of detailed county characteristics on a monthly basis over two full business cycles. Yet the more critical question was what exactly did he want to test? For all of their limitations, economic models usefully focus one’s thinking on the key parameters and potential outcomes, allowing for a reasoned series of empirical inquiries. This is where Andy most fundamentally helped Stephan learn the broader benefits of in-depth case studies; Andy’s 2010 SRSA Fellows Address in fact specifically notes that “theory development” is one of the most valuable products of good case study research.

Stephan began to understand the invaluable contributions of such tales as he slowly built his own intuition about the underlying forces at work in West Virginia labor markets. So, Stephan and his well-worn VW Rabbit went on the road, from Wheeling’s steel mill in the north to the coal mines of Williamson along the Kentucky border to the hardwood and furniture counties of the east. And slowly, the model developed through this collage of interviews, people, and regional histories. Andy, meanwhile, always enthusiastically awaited Stephan's return to hear the tales and toss in a few more of his own.

Stephan's dissertation came to fruition back at Berkeley rather quickly thereafter. Stephan's advisors, Bill Dickens, George Akerlof, and David Romer, were all especially supportive of this unusual brand of dissertation for the economics discipline, combining case study interviews, a theoretical model, and an in-depth panel study of labor market dynamics in West Virginia counties. In that sense as well, Andy’s spirit was woven throughout the final product and its published offspring (e.g. Weiler 1997, 2000, 2001), reflecting Andy’s commitment to the synergies between various forms of research.

Andy’s belief in Stephan's work continued during a brief postdoctoral research assistant professorship at WVU, which allowed him to consider and explore the various directions his research interests might take him. Stephan's home since then has been Colorado State University, where he has been happily exploring and extending the possibilities for a regional economist at the land-grant institution of a state with more advantages than most states can dream of. Stephan's commitment to neglected populations continues, a cause that Andy both cherished and implicitly taught. Indeed, farmers’ struggles in the southern part of the state and some of inner-city Denver’s problems seem more incongruous in a state with this much natural and human wealth.

Even Stephan's more recent syntheses using the concept of geographic informational asymmetries, building upon the Nobel-winning work of George Akerlof, have distinct strands traceable back to Andy. Andy’s work was all about informing the wider world, from the scholarly community to policy makers to community groups, about the parts of the world that were in shadow, the rural hamlets that usually got shunted into a residual category of some sort versus the brand-name highlights that a big city label provides (Weiler, Hoag, and Fan 2006). Andy implicitly understood that such informational imbalances could effectively cement rural areas’ economic trajectories and worked hard to balance that perspective.

Lessons on the Value of Tales and Measures Together

Stephan noted that Andy’s spirit was woven throughout his dissertation and it seems that Sarah's own dissertation, “Defining and Measuring Entrepreneurship,” was just as much in Andy’s spirit as was Stephan’s “Tales from the West Virginia Hollows.” Both dissertations drew from Andy’s encouragement to use the best quantitative measures with an equal commitment to qualitatively understanding what was going on in the world that we tried to describe.

The first time Sarah attended regional science meetings, she got to know Andy. Sarah heard him speak on economic and demographic data needs for rural America and was immediately attracted by his enthusiasm for rural development, her own being motivated by childhood experiences and a drive to make rural places better. Later, when Sarah decided to pursue a PhD, Andy was one of the first people she talked to. At the Western Regional Science Association meetings Andy introduced Sarah to his current students and encouraged her to come visit him and his colleagues at the University of Illinois at Urbana–Champaign. Sarah enjoyed that visit and soon took up residence in the interdisciplinary Regional Economics and Public Policy (REAP) student offices.

Sarah began to develop the art of storytelling, the science of better measurement, and the insights and principles that guide her research today. As she learned technical skills like input–output and spatial econometrics in class, Andy helped her translate technical analyses into stories that appealed to a variety of audiences. Soon, Sarah was explaining the local economic impact of an ethanol plant to academics, the media, and practitioners at the United County Council of Illinois.

While at Illinois, Sarah had wanted Andy to help her improve her writing; and in this regard, Andy and Sarah worked on more thorough, polished stories. Andy challenged her to produce one excellent manuscript each year rather than a higher number of more marginal, outputs—something Stephan had also instilled upon her at the Fed. Sarah thought Andy’s tendency for perfection in writing, however, might prevent the two of them from ever publishing a jointly authored article. When their first submission was accepted by the target journal, without revisions, Sarah began to realize—Andy might be on to something.

At the beginning of her last year in graduate school, Andy and Sarah went to Washington, him on a part-time sabbatical and her as an intern, both in the Farm and Rural Business branch of the USDA’s Economic Research Service. They traded walks around campus for an evening walk from the office toward home, Sarah's apartment being only a few blocks from Andy's sons’ apartment, where he stayed. While in Washington, Andy made sure Sarah got connected to the people and the data she needed, both for her dissertation but also the beginning of a career. During that fall semester, Sarah made tremendous progress on her dissertation and was able to draw on many of the lessons she had learned from Andy already.

Sarah's dissertation was an exercise in becoming a better storyteller and more effective communicator—but also, in better appreciating measures used by researchers, practitioners, and policy makers. Upon arrival in Champaign–Urbana, Sarah had planned to write a dissertation using entrepreneurship measures developed with Stephan and Jason Henderson, while all three of them were at the Federal Reserve (Low, Henderson, and Weiler 2005). Andy, however, did not buy those measures. As a prerequisite to using them in the dissertation, Sarah was charged with convincing Andy of why those entrepreneurship measures were the best available. Then she could use the measures in her dissertation. Well, as many people know, convincing Andy is not easy. In the process of convincing Andy, Sarah became unconvinced. Were these the best measures? Not just the best measures available? Somehow, Andy made her realize that she wanted to learn more about measuring entrepreneurship, about the good and bad, and to discover alternatives. This curiosity was, perhaps, one of the most important things Sarah learned from Andy.

Andy on Definitions and Measures

As Andy’s students, he taught us the importance of measures by example. One lesson was to always be forthcoming with problems in our data and to seek to improve the data. In his own index of county prosperity (Figure 1), Andy acknowledged the problems created by fuzzy concepts and choices constrained by data and politics. He developed the prosperity index using four variables that are often policy targets: poverty, unemployment, high school dropout rates, and housing problems (Isserman 2005; Isserman, Feser, and Warren 2009). Andy did not consider income as a metric of prosperity, however, because it does not account for cost of living differences, although he acknowledged poverty has a similar problem. Perhaps the lack of an optimal measure of poverty encouraged him to create an index to proxy for prosperity in the first place?

OPEN IN VIEWER

Figure 1.

Prosperity scale by county, 2000, (4 = prosperous), from Isserman, Feser, and Warren (2009)

Measuring Entrepreneurship

After questioning Sarah’s use of second-best entrepreneurship metrics while at the Federal Reserve, Andy encouraged her to think about how to define entrepreneurship, theoretically. With Andy, Sarah subsequently devised a way to create an empirical metric of entrepreneurship that better approximated the theoretical concept, which included innovation. The resulting metric is a novel combination of innovative industries and county-level entrepreneurship proxies (single-unit employer establishment births and nonemployer establishments, a subset of the self-employed) and incorporates three widely recognized functions of entrepreneurship, ownership/operation of a firm, risk bearing, and innovation. In testing the new metric, dubbed Innovative Entrepreneurship, they discovered how different it was from widely used entrepreneurship metrics that did not incorporate innovation. Figure 2 illustrates that the nonemployer establishment rate (left) and the nonemployer establishment rate in innovative industries (right) differ; indeed, they are uncorrelated and did not exhibit spatial autocorrelation.

OPEN IN VIEWER

Figure 2.

Nonemployer establishments over employment, U.S. Census Bureau, 2000 (left) and nonemployer-based innovative entrepreneurship, 2000 (right), from Low (2009).

When included in a standard model of growth in population and employment over the 2001–2007 business cycle, nonemployer-based innovative entrepreneurship had a small positive and significant (p < .01) coefficient (Low, 2009). Conversely, the nonfarm proprietorship rate, a widely used county-level proxy for entrepreneurship and self-employment, had a positive relationship with only employment growth and the microenterprise rate, the rate of establishments with 1–4 paid employees, did not have a positive relationship with either growth metric (Table 1). None of these entrepreneurship proxies had a statistically significant coefficient when regressed on per capita income growth, perhaps because of its small real growth during this period. For covariates, estimation details, and a full set of results, see Low (2009). Results suggest innovative entrepreneurship is a different, and perhaps superior, proxy for measuring regional entrepreneurship. These findings illustrate why Andy encouraged Sarah to think about definitions of entrepreneurship before pursuing her dissertation.

OPEN IN VIEWER

Table 1.

Entrepreneurship Proxies Regressed on Population and Employment Growth, from Low and Isserman (Under Review)

Y/Entrepreneurship		IE_nonemp	Proprietors	Microenterprises
Population growth		6.322*** (0.716)	−0.045*** (0.016)	−0.106*** (0.019)
	Pseudo R 2	0.69	0.68	0.68
	Log Likelihood	4,640.5	4,605.8	4,617.0
Employment growth		3.374*** (1.173)	0.264*** (0.025)	0.030 (0.031)
	Pseudo R 2	0.51	0.53	0.51
	Log L	3,171.6	3,221.0	3,168.0

*p < .10. **p < .05. ***p < .01.

In her dissertation, Sarah found that the drivers of nonemployer-based innovative entrepreneurship differ from the drivers of the self-employment rate. For example, when using growth in self-employment as a dependent variable, the coefficient on the ERS natural amenity scale (McGranahan 1999) was negative and significant (p < .01), suggesting self-employment growth is higher in low-amenity areas; the opposite result holds when growth in innovative entrepreneurship was the dependent variable (Low 2009). This finding illustrates how different definitions and measures can identify different economic development and policy strategies as drivers of growth, and another reason Andy encouraged thoughtful definitions to be used in regional science research.

Defining Rural and Urban

In his 2005 article, In the National Interest: Defining Rural and Urban Correctly in Research and Public Policy, Andy stressed that most counties are combinations of urban and rural areas (p. 494), yet widely used definitions and measures of rural are available only at the county level (Isserman 2005). Also, Andy noted that the Office of Management and Budget (OMB) definitions of metropolitan and nonmetropolitan are widely misinterpreted as urban and rural. He warned against referring to metro counties as urban and all other counties as rural, noting that this practice persists because no one has identified a better alternative for studying rural within the federal statistical system. Definitions of rural are so varied that anywhere from 58% of the U.S. population to a mere 2% is rural (Isserman 2007).

Andy argued that the OMB typology does not ascertain the rural character of a county well, nor was it designed to. Most counties are a mix of rural and urban, so metropolitan statistical areas, often composed of urban-character counties, that is more urban than rural, and surrounding rural-character counties, are not a good substitute for urban.

Andy argued that the rural character of a county should be used to measure and define rural. By combining the OMB county-level designations with the Census Bureau’s definition of rural and urban, which is at the block level, Andy created a rural–urban density typology (Figure 3). The rural–urban typology incorporates block-level data into a county-level rural–urban character typology, positing urban, mixed-urban, mixed rural, and rural counties, as described in Isserman (2005). The typology can then be used with the wide array of county-level data available but allows for differing degrees of rural/urban within a county. It is research like this that has the potential to clarify definitions and continue to reshape our understandings of place.

OPEN IN VIEWER

Figure 3.

Rural–urban character, by county, 2000, from Isserman (2005).

By combining OMB and Census definitions to create a new way to define rural and urban, Andy learned that rural counties within metro areas do better than rural counties within nonmetro areas. Indeed, recent work at USDA’s Economic Research Service has studied the dichotomous nature of nonmetro counties, in the context of out-migration counties and those without high out-migration (McGranahan, Cromartie, and Wojan 2010). These authors find the differences among nonmetropolitan counties are stark and relate to urban connectivity.

The Definitions of Regions also Affect Measurement

Andy also exhibited interest in defining regions correctly (Isserman 2002, 2008). In the earliest example, we could find of his work on defining regions, a 2002 proceedings paper from a conference on the New Power of Regions, convened by the Kansas City Fed’s Center for the Study of Rural America, Andy wrote about self-defined regions (those springing-up around a business opportunity), economic regions (defined around the basis of their economy), and natural resource regions (shared important natural resource, such as farm land); also, Federal action regions, and state and local action regions. He noted that the use of counties as building blocks for regions could create oddities due to counties’ combined rural and urban character and called for the major federal statistical agencies to process and make-available urban and rural data for each county.

Andy further developed his ideas about defining rural regions during Sarah’s years at Illinois with geographic information system (GIS) help from the REAP grad students; this work manifested in a presentation at the annual meetings of the Association of Public Data Users’ conference in 2008. During this presentation, a refined look at defining regions, Andy argued that, “routine use of areal units larger than counties and smaller than states would sharpen and expand our knowledge of regional problems.” He argued that the definitions of regions used in analysis is important and should be tailored to the analysis at hand, rather than based on political boundaries and other arbitrary definitions of regions. Since the advent of county-level data, it seems as though most rural and regional work is done at the county level or by aggregating counties. Although counties can be a useful unit of analysis, Andy believed that researchers need the flexibility to be able to create their own functional regions with publicly available data.