Why Regions? Why Now? Who Cares?

ABSTRACT:This article examines the emergence of the region as a platform for achieving economic competitiveness and social equity. We use a combination of empirical exercises to examine the timing and degree of actor interest, including convergence regressions and coefficients of variation analysis to see if regional economic and social performance is more heterogeneous, and logistical analysis to determine what metropolitan-level factors can explain the rise of regionalist interest by business organizations and/or social justice organizations. We find that performance is, in fact, generally more heterogeneous, that different actors are motivated by different factors, and that interest by business groups seems to trigger a "reactive regionalism" by equity proponents (but not the other way around). Notes1 We should note that specifying the exact boundaries of the regional scale is itself a "subjective" question—some regional leaders may define the region to include many metropolitan areas (as with the business-oriented Bay Area Council in California or the philanthropic multimetro Fund for Our Economic Future in Northeast Ohio). To keep the empirics tractable, our tests are based on one standard definition of regions: the current metropolitan area definitions set by the OMB under the rubric of core-based statistical areas (CBSAs) but we acknowledge that the very definition of a region is up for grabs in the new regional framework.2 See CitationGereffi and Korzeniewicz (1994) and CitationHenderson, Dicken, Hess, Coe, and Wai-Chung Yeung (2002). There is a related argument that large global city-regions are the optimal location for the management of increasingly complex global production networks (CitationSassen, 1991; CitationScott, 2001). We suspect that this argument is correct but its empirical implication is mostly that some regions are becoming more important than others; our focus here is whether the regional level matters more for most metropolitan economies.3 Scholars from a variety of disciplines have examined the economic transition to the Information Economy and its spatial implications. Manuel CitationCastells (1996) provides the most complete explanation.4 There were also a set of authors, including CitationSavitch, Collins, Sanders, and Markham (1993) and CitationVoith (1992), who argued for a complementarity of city and suburban incomes and offered supportive empirical evidence. CitationHill, Wolman, and Ford (1995) provided an early critique of this complementarity hypothesis, raising important issues of simultaneity that both CitationPastor et al. (2000) and CitationVoith (1998) sought to address.5 For a less simplistic approach, see the literature on endogenous growth theory such as CitationRomer (1990). In this framework, increasing returns, technological change, and other factors matter greatly.6 There are some complexities as to whether the traditional neo-classical growth model—which assumes a closed economy and investment from local savings—is truly applicable but the heuristic implication is that of faster convergence, something supported by the data. See CitationHiggins, Levy, & Young (2006) for a discussion of this issue.7 Previous conditional convergence studies include variables that measure the relative size of racial and ethnic minority populations. We tested models that included both the percent Latino and percent African American but found a high level of collinearity with the percent foreign born in the case of percent Latino, and a somewhat unstable effect of percent African American between the two time periods. Since the inclusion of such variables had less purely economic rationale and had little impact on the central finding regarding the change in the beta coefficient on initial per capita income, we instead present the most parsimonious model with consistent variables in each decade.8 See CitationStorper's (1997) discussion of "untraded interdependencies"—formal associations and informal networks in which information, business tips, and technological innovations are exchanged. Such factors do not show up easily in multiregion databases, and a declining significance of other more traditional factors can signal a rising significance of these regionalist "intangibles."9 Note that for all income measures derived from the decennial census, the data are for the years 1979, 1989, and 1999 since measures of income are taken in the census from the previous year.10 At least one reader of an early draft of this work expressed concern that the particular definition of metropolitan areas used for this investigation—the CBSA—was too geographically specific in the larger metro areas because it sometimes separated out what was formerly considered a single "Consolidated Metropolitan Statistical Area (CMSA)" under the 1999 Census definitions into two or more CBSAs under the 2003 Census definitions. This, it was suggested, allowed some large regions to be represented as more than a single observation and thus carry more than their share of influence over the CoV calculations. Due to this concern, we made a separate set of all CoV calculations in which we compared the experience of those CBSAs falling inside what was formerly a CMSA (of which there were 17 in 1999) to those falling outside such an area, and found that the dispersion in our measures of economic growth and equity over the period was even more pronounced for those falling inside than for those falling outside a CMSA—a result that held (and even became more pronounced) when attention was restricted to the top 100, top 50, and top 25 CBSAs. The indication, we think, is that there has been dispersion in outcomes within the CMSAs themselves. This finding was supported when we inspected the variation in our measures across the CBSAs falling within several particular CMSAs over the two decades, and that result is supportive of the use of the CBSA to define regions given such observed heterogeneity within CMSAs.11 As a comparison, we also calculated coefficients of variation for earnings and employment at the state level—both for all states, and for those containing any CBSA and any CBSA of the particular size categories broken out above—and found similar, but less pronounced results. This suggests that it is the metropolitan areas (in this case measured by CBSAs) rather than states as a whole that are driving the dispersion of outcomes that has been observed over the past two decades—and this is why organizing and intervention at the metropolitan level, albeit supplemented by state policy changes, might be useful.12 The model is akin to the income convergence model presented earlier; it is of similar design to that found in CitationCrandall and Weber (2004), but uses regional rather than census tract-level measures.13 This is perhaps due to collinearity with the single-parent-family and percentage foreign-born measures, whose coefficient rises sharply in the 1980s as compared to the 1970s while the other covariates remain relatively stable.14 We tested out other formations of the polychotomous response variables and results were similar. This specification was chosen because it seemed to best capture regional interest relative to the size of the region (or CBSA), which can vary greatly.15 For each regression, we also inspected the results obtained when the particular regions hosting the PolicyLink regional equity summits and ARS conferences were omitted, out of concern that they might have representation or stronger representation simply because such meetings would be so much more convenient to attend for people living in the hosting region. All such results are essentially the same as those reported in Tables 7 and 8, so we stuck with reporting the results for the full sample of regions.16 We tested other specifications of business or equity attendance, including simple dummy and more complex categorical variables. We also conducted a residual analysis to see, for example, whether business interest greater than that predicted by business attendance regression was correlated with attendance by equity proponents at the PolicyLink conference; the idea here was that this was a very good signal of active regional agenda-setting by the business class. All these formulations yield the same result as that obtained with the simpler measure in the text—equity proponents are motivated by business attendance but not the other way around—and so we stick in the text with the more straightforward approach.17 The marginal effects were calculated using a standard formula that is perhaps the most common approach to estimating marginal effects: . In this formula, β is the estimated logit regression coefficient for the explanatory variable and is the mean value of the response variable (or the percentage of cases that take on a value of one). Also, note that the number of observations included in the regression (which is the same for all logits reported), 187, is five short of our overall sample of CBSAs considered (192). This is because a few variables—one from the regional equity logits and one from the ARS logits—are not observed for all CBSAs in our sample. To make the set of CBSAs in the regressions consistent across all specifications, we restricted the sample to the set of CBSAs with observed values for all variables in all specifications. We also tested the separate models without placing such a restriction on the sample and the results were essentially the same.18 We thank an anonymous referee for pointing us in this direction by noting the emphasis of equity proponents in general and PolicyLink in particular on housing. The measure we used is figured as the ratio of the share of central city renters spending more than 30% of household income on rent to the suburban share. Other measures of straightforward rental unaffordability were tested (such as the regional and central city percentages of renter-occupied households spending more than 30% of income on rent) were not as significant as the relative measure—but that makes sense since the summits are about regional equity and not simply a shared regional burden of high housing costs, something that seems to drive business concerns. It may be useful to note that introduction of this measure reduces the significance of the income distribution measure slightly, a pattern that makes sense given that this variable reflects a geographic dimension of inequity.19 This, for example, was the case with the Social Equity Caucus (SEC) in the Bay Area which was originally created by an environmental justice intermediary, Urban Habitat, as a way of responding to a series of "sustainability" conversations sponsored by a business organization, the Bay Area Council. The SEC subsequently evolved into a fuller coalition with its own agenda, including advocacy for mass transit throughout the San Francisco region and organizing for better urban planning in distressed areas like Richmond, California. "Reactive regionalism" is not confined to simply reacting but can develop an independent program; we are simply trying to see which sector tends to lead the other.20 Out of concern that the measure capturing the percentage of workers working outside their county of residence was related to the geographic size of counties—which tend to be much larger in the western states of the nation than the remainder (particularly the South and East)—we inspected its mean value across the four major regions of the U.S. delineated by the Census. We found the mean value for CBSAs in the West of 15.65% was not too different from those in the other four regions, which it ranged from 20.82% to 22.46%, so we kept this measure in the final model. We are also reassured by the fact that the reported specifications include broad regional dummy variables for three of the four regions, which should capture the differences in county size since they are quite strongly related to the broad census regions.21 The definition of "high-tech" occupations was borrowed from the 6-digit NAICS-based definition specified by the American Electronics Association (AEA). See http://www.aeanet.org/Publications/idmk_naics_pdf.asp.22 On the other hand, the ARS attendees were fewer than the PolicyLink attendees altogether and so the breaks were tighter; as a result, perhaps we are simply trying to tease too much significance from too little variation in the data, and the dichotomous logistic results are more sensible.