The Geography of the New Economy
R. D. Norton

The ultimate irony in the placeless world is that some places organize the rest.
Manuel Castells 1998, p. 188

In 1967, at the crest of the Old Economy’s development, John Kenneth Galbraith declared the individual entrepreneur obsolete, saying “only the group has the information that decision requires” (Galbraith 1967, 1985, p. 104). Today, in the light of history, we see things differently. In retrospect, “the group” in the traditional managerial corporation looks more like a stultifying bureaucracy, where the safest tactic was the non-decision. (Larry Farrell 1993.)

By the same token, the I.T. case study in Part B was intended to show the vital role played by newcomers, acting as entrepreneurs, to overthrow the established order and blast through the tendencies toward stagnation that past success seems to breed. From that standpoint, the difference between the U.S. economy, on the one hand, and those of Japan or France or Germany, on the other, has seemed to lie in the superior opportunities the U.S. has afforded newcomers—geeks, freaks, immigrants, and other outsiders.

Yet the basis for Professor Galbraith’s verdict remains of interest. New technologies are not necessarily easier to understand today than in 1967. Group (or, in today’s parlance, “team”) cooperation, consulting, and coordination are often as crucial to product development and innovation today as they were then.

What has changed, it would seem, is the legitimacy of hierarchy. A primary lesson of the last third of the 20th century was that hierarchy is antithetical to the free and open flow of strategic information, “the information that decision requires.” This, as many people have observed, is the impression one might glean from the fates of the U.S.S.R. and of U.S. corporate dinosaurs (like Sears or General Motors) alike.

More recently, of course, the proliferation of computer networks both within and between organizations has also made hierarchy less tenable. As a result, the 1990s saw powerful tendencies toward flatter organizations; burgeoning alliances between large and small firms; and deepening networks between firms and venture capitalists, universities, and governments.

Information flows remain vital, in other words. But now PC networks and spatial proximity provide increasingly complementary channels for the horizontal transmission of strategic information.  One result, as manifested most vividly in the U.S. in perhaps a dozen large and mid-sized cities, is a new system of innovation, driven by partnerships between knowledge workers and venture capitalists.

That is our current point of departure.

The purpose of this section, then, is to consider which large American cities (more specifically, metropolitan areas) are spearheading the New Economy’s next round of development—and to ask how they have emerged as centers of innovation.

Naturally, any such inquiry needs to begin with a deep bow to Silicon Valley.


In "The Valley of Money's Delight" (The Economist, March 29, 1997), John Micklethwait cites economic cultures as the catalysts that determine whether networks communicate. As he observes, "Research has increasingly concentrated on clusters—places (such as Hollywood or Silicon Valley) or communities (such as the overseas Chinese) where there is 'something in the air' that encourages risk-taking."

He lists 10 features of Silicon Valley's economic culture that help explain the area's dynamism:

  1. Tolerance of failure
  2. Tolerance of treachery.
  3. Risk-seeking
  4. Reinvestment in the community.
  5. Enthusiasm for change.
  6. Promotion on merit.
  7. Obsession with the product.
  8. Collaboration.
  9. Variety.
  10. Anybody can play.

This list points up the fluidity of the Valley as an economic environment. One of the qualities it conveys is a sense of loyalty to the place, rather than to the firm. By extension, it suggests a milieu conducive to spin-offs and start-ups—an environment that can be termed “Economy 2.” (Martin Kenney and Urs Von Burg 1999.)

Linked background sketches on cluster theory offer further observations on the connection between information flows and spatial access. A first module surveys neoclassical approaches to cluster theory, those focusing mainly on spatial externalities. The second, on what I term post-neoclassical models, considers network-based industrial systems, path dependency, increasing returns, and dynamic agglomeration economies. As a reminder that any such hard-and-fast dichotomy between neoclassical and post-neoclassical views is fraught with peril, a sidebar locates cluster theory within the spectrum of urban growth paradigms. A useful set of links on cluster-based state economic development policies appears in the Hubert H. Humphrey Institute’s list at the University of Minnesota.

That said, we can turn directly to a range of diverse views as to the role of large cities in the American economy. The goal is to discern which specific cities are best positioned at the Millennium to facilitate the information flows likely to promote innovation.


A few framing points about the U.S. system of cities can be offered now. The unifying theme is the role of history as help or hindrance to a metropolitan area’s economic performance. From the standpoint of evolutionary economics, this is an issue of path dependence. From the standpoint of cluster theory, it overlaps with the question of specialization vs. diversity.

(1) City Roles in the World Economy

In a conference announcement from the University of Newcastle (England) in 1998, the organizers proposed a typology of cities based on 10 distinct city types. The conference theme was “Cities in the Global Information Society,” so the taxonomy can be understood in that light. Here is the list, along with examples suggested by the organizers:

  1. Old-industrial
    (e.g., Newcastle, Pittsburgh, Essen)
  2. Global
    (London, New York, Tokyo, Singapore)
  3. 2nd Tier regional and national capitals
    (Amsterdam, Dublin, Milan, Taipei, Toronto, Sydney)
  4. Newly-industrializing
    (Pearl River Delta)
  5. Former communist
    (Moscow, Warsaw, Budapest)
  6. Globally marginalized
    (Soweto, sub-Saharan Africa generally)
  7. Information-processing
    (Sunderland, Bangalore, Kingston [Jamaica])
  8. Resorts and tourism
    (Palma, Orlando)
  9. Logistics
    (North Carolina [sic], Rotterdam)
  10. New planned
    (Malaysia's Multimedia Corridor and Japan's technopoles)

With a couple of obvious modifications, a similar taxonomy could be applied to the U.S. system of cities, using, say, categories 1-3 and 7-9.

In particular, asking which of America’s largest cities are “industrial” in origin (type 1) is a fruitful exercise.

(2) American Metropolitan Evolution (Revisited)

For example, TABLE 10 links changes in manufacturing employment after 1970 to the mid-century industrial legacies of 30 large U.S. areas.  It reveals a record of large losses by industrially specialized areas.

The 30 metro areas contained the 30 largest cities in 1970, when large-scale losses of manufacturing jobs were about to begin. The areas are ranked by their population sizes in 1910, at the end of the nation’s heavy industrialization and before the automobile or electricity had had much impact. This historical approach (introduced in Norton 1979) owes much to the geographer John R. Borchert’s proposed sequence of technology epochs in a classic 1967 article, “American Metropolitan Evolution.”

The dozen areas that had reached the largest size by 1910 can be termed “industrial.” The dozen then smallest were deemed “young.” In between, such areas as Los Angeles, Washington, D.C. and Seattle are “anomalous,” in that much of their growth had occurred after 1910 but before 1950. Among the variables that then align by age-class are (1) population density, (2) industrial structure, and (3) unionization rates. (Norton 1979.)

Regionally, 11 of the industrial areas were in the Manufacturing Belt, and 10 of the younger areas outside it. (See MAP 2, which is adapted from Norton 1979, p 25.) At mid-century, the dozen industrial areas still had an average 35 percent of their 1950 workforces in manufacturing jobs. In contrast, the dozen termed younger had an average of only 19 percent.

Their roles as exporters of industrial goods to the rest of the U.S. and abroad left the mature metro areas vulnerable to huge losses in manufacturing employment after 1970. The combined manufacturing job losses from four of them—New York (down 658,000), Chicago (326,000), Philadelphia (277,000), and Detroit (198,000)—exceeded the entire U.S. loss (1,395,000). Most younger areas added manufacturing jobs over the period, including a few (Atlanta, Dallas-Fort Worth, Houston, San Diego, and Phoenix) with sizable absolute gains.

As to changes in total employment, the contrasts between industrial and younger areas are milder, but still pervasive. The U.S. added 55 million payroll jobs from 1970 to 1998, for a percentage gain of 78 percent. Relative to this national rate, three points about the 30 areas might be made:

  • A few industrial areas (New York, Pittsburgh, Cleveland, Buffalo) had extremely low job growth—below 10 percent.
  • The median figure for the 11 older areas in the Manufacturing Belt, 35 percent (for Chicago), was less than half the U.S. rate.
  • The median for the 10 younger areas of the South and West was 157 percent, twice the national rate. Atlanta, Dallas, Houston, and Phoenix added more than 1 million jobs (as did a now resurgent Chicago, Los Angeles, and Washington).

Even in the late 1990s, with brisk job growth nationwide, the older areas still lagged. As FIGURES 14 and15 document, aggregate job growth from 1995 to 1998 remained only about half the rate in most older areas as in most younger ones.

In sum, the specialized industrial roles of the mature areas led to large-scale losses of manufacturing and sluggish growth in total employment. While not exactly news, this remains point number one in any overview of the system of U.S. cities.

From the standpoint of cluster theory, we might put all this a different way: Specialization can be good for city growth—or not! It all depends on the nature of the activity, the pattern of demand from the rest of the world, and the chemistry between the activity and “learning” on the part of the city’s workforce and knowledge base.

(3) The Ladder of Influence

An opposite view comes from David Warsh, who writes an economics column for the Boston Globe. Prompted by the purchase of the Los Angeles Times by the Chicago Tribune in early 2000, Warsh proposed an informal ranking of the leading centers of U.S. influence. His admittedly impressionistic list refers to “education, finance, and media industries…and the capacity to absorb the latest streams of immigration….” (Warsh 2000, p. E1.) By this reckoning, the three largest cities, New York, Chicago, and Los Angeles, are also the three most influential, the places where U.S. opinions and attitudes are shaped.

Then there are “the other American cities of international importance—Washington, D.C., Boston, Miami, San Francisco and, possibly, Seattle…world centers in certain fields.” In this reckoning, Washington qualifies only because it is the capital. Boston and San Francisco make the top 8 by virtue of their financial and university strength. Miami qualifies as the gateway to Latin America and the Caribbean, and Seattle as a “high-tech nursery.” Global cities in specialized realms, these five fall just below the top three, New York, Chicago, and Los Angeles.

Warsh’s conclusion? “This is not to rob a dozen other U.S. cities of their significance. …But the hierarchy is well-established, and here, as in Europe, the oldest cities tend to remain at the top.” (Warsh, p. E1, emphasis added.)

This curious generalization may have some relevance to media and entertainment. But it completely misses the dynamic of renewal by which younger centers have restored the American economy to global leadership. A sense of that dynamic can be seen in the recent upheavals in the U.S. system of cities—indexed not only by job growth, but also migration choices and I.T. roles.

(4) Tech-Poles: The Milken Institute List

Consider, for example, the Milken Institute’s 1999 ranking of “Tech-Poles.” These are the U.S. metropolitan areas that stand out by virtue of their size and specialization in a broad range of high-tech activities. (DeVol 1999, p. 67.) When an area’s percentage share of U.S. high-tech output is multiplied by its high-tech output location quotient, the result finds San Jose (Silicon Valley) the runaway leader, followed by Dallas, Los Angeles, Boston, and Seattle. The next five are Washington, D.C., Albuquerque, Chicago, New York, and Atlanta.

In other words, four of the top five areas are from the South and West, as are three of the next five (once we recognize that the Washington, D.C., area’s high-tech center of gravity is northern Virginia). That adds up to six of the top seven metropolitan areas from the South and West, as measured across the gamut of high-tech activities.

A still sharper regional watershed can be seen for domestic migration.

(5) Magnet Metros: The Seattle-Atlanta Line

Niles Hansen contends that domestic migration flows in the 1990s give a clean read on the economic opportunities offered by major metropolitan areas. (Hansen 2000.) In part this view is based on the observation by Glaeser that domestic migration flows offer a better indicator of an area’s success than per capita income growth, because the latter may include a “bribe” component in wages to offset urban disamenities.

The areas with the largest 1990-1997 in-flows can be found below what Hansen terms the Seattle-Atlanta line. (See MAP 3, which is based on data presented in Hansen 2000, Table 2.) The numbers range from over 300,000 in Atlanta, Phoenix, and Las Vegas to gains between 120,000 and 162,000 in Seattle, Portland, Dallas, Denver, Austin, Raleigh (which is just north of the line), and Orlando.

In terms of size, all the “magnet metros” had fewer than five million residents in 1995. The largest was Dallas, with 4.7 million resident. The next largest were Atlanta, 3.6 million, Seattle, 3.4, and Denver, 2.3. That meant that no magnet metro was as large as any of the 8 largest areas: New York, Los Angeles, Chicago, Washington, San Francisco, Philadelphia, Boston, and Detroit. Each of these 8 largest had over five million people, and each had net domestic outflows.

The map serves as a visual reminder that size is but one of several linked variables. It portrays a regional realignment from high-density, high-cost, older areas in the Manufacturing Belt and California to younger, low-density centers. In turn, this shows up in the data as a move from larger to smaller cities.

The net effect, Hansen concludes, “has been a definite shift downward in the urban hierarchy in terms of where Americans want to live and work.” (Hansen 2000, p. 12.) And as he demonstrates, the shift is not only from the largest to mid-sized metros, but also from the Manufacturing Belt and California to younger areas in the diagonal band between them.

(6) The Perils of Specialization, Continued: I.T. Hardware

Just as specialized roles proved a heavy load for industrial metros after 1970, so too did high profiles in computer production and electronics between 1986 and 1996. (TABLE 11.) The precipitants were declining U.S. employment in computer production (SIC 357), slow job gains in electronics (SIC 367), and decentralization of both to rural states.

Thus the three areas with the greatest initial specialization in computer and electronics production accounted for over half of all hardware jobs lost nationwide from 1986 to 1996. (See Equation 1.) Phoenix, Boston, and Los Angeles combined for hardware losses of 73,000 jobs.

How different is the lesson here from that of the de-industrializing “industrial” cities after 1970? The two cases seem closely related, and not only because the I.T. hardware losses are one component of the larger losses in manufacturing employment in older areas. In each story, initial production centers specialized in sectors that would add little or no employment nationally, a scenario that tends to be accompanied by rapid dispersal to competing domestic sites, including non-metro locations.

Put the other way around, one of the ways the U.S. as a geographical entity retains employment relative to offshore locations is by offering both competing centers of innovation and lower-cost (including non-metropolitan) environments.

In any case hardware was only half the story of metro I.T. growth during “the break-up of the old computer industry” (Grove 1993, p. 57).

(7) In Sum: Diversity and Adaptive Capacity

We are exploring the geographical origins of the New Economy in the U.S. Regionally, the PC revolution had largely western coordinates, as Part B showed. As the New Economy moved into high gear in the mid-1990s around the Internet, a different geographical logic took over. The underlying forces shaping place competition increasingly came to include media and finance, not just I.T. Accordingly, the creation of technology-based start-ups would now depend on resources available to a few of the most diversified of the faded industrial centers, the industrial cities identified in Map 2.

From either perspective, that of the U.S. resurgence during the PC revolution or the Internet explosion of the late 1990s, the diversity of the system of cities may well have added to the U.S. economy’s adaptive capacity.  As Clem Tisdell observes, “Industrial diversity (more generally diversity of driving attributes in dynamic systems) can have value in increasing the likelihood that an economy (or system) can jump to a superior state.” (Tisdell 1999, p. 163.)

>By extension, we might surmise that continentality and regional diversity aided the U.S. immeasurably in its shift from mature industries and cumbersome managerial corporations to new forms and sources of growth.


Now we come to the third basic tendency transforming the U.S. system of cities. The first point has been the influence of history—known in the cluster literature as path-dependence—in the economic performance of the 30 large areas we are monitoring. The second is domestic migration flows, as shaped by the influences that make mid-sized younger centers from Seattle to Atlanta “magnet metros.” The third is the agglomeration of knowledge workers in a dozen or so major areas blessed by a favorable mix of venture capital firms, universities and research institutes, and a crackling atmosphere—typically a high-amenity location where ideas and possibilities are, in Alfred Marshall’s term, “in the air.”

Where, then, did concentrations of I.T. workers grow most rapidly between the mid-1980s and the mid-1990s? In light of the stagnant job growth in computer and electronics hardware employment, the answer turns largely on software and other computer services.

We will find that the geography of job growth in software had a logic opposite that of hardware. That is, the places that specialized most in software and other computer services in the mid-1980s would then go on to record the largest software job gains over the next decade. Since software and other computer services added jobs at a rapid clip during this interval, for most of the 30 areas the employment gains easily outweighed computer hardware losses. (Exceptions were two hardware centers, Los Angeles and Phoenix.)

In turn, some of the initially specialized areas saw software expansion interact with the local venture-capital base to spur new technology-based business creation, as measured by initial public offerings (IPOs). The result for Boston and New York, “industrial cities” in terms of the timing of their industrialization, has been a dramatic comeback in the innovation race, fueled in good part by specializations in higher education, finance, and media.

Hence the spatial chemistry for innovation to be documented now. The indicator to be used is the IPO, the issuance by a privately held company of common stock to the general public. While 600,000-800,000 new businesses are formed each year in the US, only about 400 companies reach the moment of an IPO. To that extent, IPOs can be viewed as survivors of a selection process to single out elite start-up companies promising investors high profits because they can do something new—Schumpeter’s touchstone for innovation.

Part Real, Part Surreal: The Internet Gold Rush

At the same time, this may be the dimension to the New Economy best described my Mark Zandi’s term, “part real, part surreal” (Zandi 1998). Realistically, an IPO can be viewed as an attempt on the part of promoters to “sell” a new idea to the investment community. During the Internet Gold Rush of 1998 and 1999 some IPOs have had more hype than content, as the shakeout of dot-com’s in April 2000 demonstrated. To that extent, IPOs are an imperfect measure of innovation—an indicator of market fads as well as of genuine new ideas.

For now, suppose we view IPOs as a rite of passage for an idea-based start-up firm, a moment of truth when the firm’s defining premise is put to the test of the market.  The question is, where are such new ideas most likely to occur, to be put into practice, and to reach the stage of going public?

As a working hypothesis, we might surmise that IPOs in the late 1990s were most frequent where knowledge workers could hook up with venture capitalists—the suppliers not only of money, but of management expertise of the kind most technology-based start-ups lack.

Accordingly, the topics to be explored now are (1) the new lineup of software centers, where I.T. workers are concentrated; (2) the prominence of venture capital (VC) firms in the 30 areas; and (3) the rate of IPOs in an area per million residents.

Software Centers

As noted, software centers had employment gains that swamped their losses in hardware. (TABLE 12.) For example, of the 30 large areas, Washington, D.C., was most specialized in SIC 737 (computer services) in 1986. Thanks in large part to the explosion of software and telecommunications in northern Virginia, the D.C. area also had the largest gain in computer services employment, over 50,000. At the other end of the spectrum, the least specialized area in 1986, New Orleans, had the smallest increase in computer services employment.

Here we have a stylized dichotomy, which in this case may be accurate. The contrast is between two southern centers, the National Capital Region (with its abundance of government agencies, including the Pentagon, the outsourced private contractors, the media covering the federal government, the spectrum of universities, and the tradition of intellectual conflict and ferment) and New Orleans—a city whose chief claim to fame is the Marti Gras. The first led the list in terms of job growth in computer services. The second came in last.

How general was this tendency? To find out, we can test changes in employment from 1986 to 1996 against initial location quotients. (A “location quotient” expresses the proportion of a sector like I.T. in a place’s total employment, relative to the same proportion for the U.S. Hence location quotients above unity would indicate that the area is more specialized in the activity than the nation as a whole.)

Equation 2 indicates that a difference of one point in 1986 location quotients between areas was associated with an increment of 10,000 computer-services jobs over the decade after 1986. Whatever the bundle of variables represented by the initial location quotients, together they account for nearly 60 percent of the variation in job gains.

In contrast to hardware jobs, then, this was an example of virtuous specialization.  In a rapidly growing employment sector nationwide, initial centers tended to grow as rapidly in percentage terms as others, hence scoring larger absolute gains.

Does Venture Capital Stay Local?

Now we come to the financing mechanism. The starting point is that lead VC firms tend to “stay local.” The reason is their need for routine face-to-face contact with supported early-stage firms. As a Silicon Valley journalist notes, “If you need to meet with a company every week or other week to get it off the ground, you don’t want to have to jump on a plane and cross three time zones to do it—especially if you generate high returns off companies based in your own proverbial backyard.” (Shawn Niedorf, “New Yorkers Not Talk of Town,” San Jose Mercury-News, on-line, March 7, 2000.)

At the same time, Niedorf’s qualifier (“especially…”) points up the key premise in her argument. What if you cannot find promising companies right in your backyard? Which comes first, the VC chicken or the start-up egg? At this point a word about the origins of VC—and its migration west—may come in handy.

Venture capital was invented in the form of Boston’s American Research and Development (ARD) at the end of World War II as a deliberate attempt to incubate new activities to offset the decline of New England’s ancient industries. By the 1960s, venture capital also took hold in Silicon Valley, where Shockley Semiconductor had enhanced the presence of Hewlett-Packard and the Stanford Research Park. Both Boston and Silicon Valley would go on to become the nation’s primary VC centers and hotbeds of technology-base start-ups.

On the other hand, New York or Chicago venture capitalists may take part in syndications, through “co-investments” with lead VC firms elsewhere—Silicon Valley, Massachusetts, or more recently Texas, for example. This was the tendency documented in a 1992 study of VC’s role in 8 major centers. The authors classified 8 VC centers as technology-oriented (Silicon Valley and Denver), financial-oriented (New York and Chicago), or hybrids of the two (Boston, Minneapolis-St. Paul, Texas, and Connecticut). (Florida and Smith 1992, p. 201.)

At that time they found that “just 7 percent of the investments made by New York venture capitalists were made in-state,” vs. 70 percent in-state in California. In between was Massachusetts, whose VC firms made 40 percent of their placements in-state, and 30 percent to California start-ups. (Florida and Smith, p. 193.) (A different angle on the feasibility of long-distance relationships, as facilitated by airline connections between emerging and established centers, appears in a recent study of innovation in Texas cities.)

Florida and Smith’s study of the 8 VC centers appeared in 1992. In the meantime some things have changed, such as the rise of New York City’s “Silicon Alley,” which specializes in media-based Internet start-ups. One might therefore expect to find deepening ties between Wall Street venture capitalists and Silicon Alley entrepreneurs.

VC "Funds" as an indicator of Local Supply

The hypothesis, then, is that the frequency of IPOs in an area will increase, the greater the supply of venture capital in the area. How, then, should be measure “supply”? Lacking more precise data, a good indicator of the size of an area’s venture capital base is the number of separate funds being maintained by the area’s VC firms. Each fund in a VC firm’s “portfolio” represents a separate sector (e.g., biotech, network software, or e-commerce). And each has a separate and finite duration (perhaps five or 10 years), to be liquidated at maturity. (TABLE 13.)

Note that this indicator measures where placements originate—not where they land. Since the purpose of a VC placement is to bring the early-stage firm to a successful IPO, linking IPOs to where placements land would be tautological, explaining nothing. (That is, when a Chicago venture capitalist has a placement in Silicon Valley, the IPO is all but certain to occur in the Valley.) In contrast, we are testing Shawn Niedorf’s maxim: lead or solo VC firms prefer, in effect, to stay home because of the need for frequent face-to-face contact with supported start-ups.

In short, the premise is that start-ups in a given metropolis are more likely to find VC financing and assistance if more VC funds are being run there.

An IPO a Day: 1996-1999

A word about the IPO data. From May 1996 to November 7, 1999, 1,532 IPOs were launched in the U.S. That averages over 400 per year, or more than one a day. The three and a half years surveyed is the interval covered by the data-base in Hoover’s on-line IPO directory ( The data-base permits counts by industry, by state, and by metropolitan area.

Over that interval from mid-1996. about three-eighths of the total count have been in some sense “digital,” linked to computing, semiconductors, software, networks, or e-commerce. (The proportion rose sharply in 1999, as the Gold Rush gathered speed, to about 60 percent.)

In absolute terms a handful of areas dominated the metro landscape for IPOs over the period from July 1997 to late October 1999. New York and Silicon Valley each were home to about 200 IPOs. Adding Los Angeles’s 94 and Route 128’s 90 gives a figure for the four top metros of over half of the 30-area total—and about 40 percent of the U.S. total (TABLE 14). Like its progenitor venture capital, IPO activity thus tends to be concentrated in a few major centers.

In addition, the large number of IPOs for the New York area suggests a sharp increase in start-up activity, triggered in part by media-linked Internet firms. No longer does money raised by venture capitalists in New York all go to other regions.

At the same time, some unexpected places also have high IPO rates, once we discount the effect of absolute population size.

IPO Rates by Area, Relative to Population

Standardized for population, how do individual areas compare to the U.S. averages, i.e., about six IPOs of all kinds, and about two “digital” IPOs, per million residents? (To repeat: digital offerings include not just Internet issues, but any that relate to computers, electronics, or software.)

For IPOs generally, the highest rate was Silicon Valley (approximated by combining the San Francisco and San Jose metropolitan areas). It had nearly 30 IPOs per million residents, about twice the rate of any other area. As in the Milken Institute ranking of high-tech output noted above, the San Jose/San Francisco region is in a class by itself.

A dozen other areas on the list came in above the U.S. average. Other entries include second-place Denver (above Boston or New York), Seattle, Atlanta, San Diego, Baltimore, and Philadelphia.

In contrast, both the smaller areas in the South and West and the more “heavy-metal” areas of the Midwest lagged the national averages.

For digital IPOs per se, the top five entries in TABLE 15, are Silicon Valley, Denver, Route 128, Seattle, and Washington, D.C. (i.e., including northern Virginia). By contrast, Philadelphia, Houston, Kansas City, and Chicago are less prominent digitally than for IPOs in general.

Digital IPO Rates as a Function of the Two Variables

Denver aside, we seem to have arrived back to a list of “the usual suspects” for clusters of innovation. To what extent does this outcome reflect the proposed explanatory variables—the supply of venture capital and the relative size of an area’s software sector? Two measures of the latter influence have been found relevant. One refers to the share of an area’s total employment in software jobs in 1996. Another refers to the rate of growth of employment in computer services between 1986 and 1996.

It turns out that three-fourths of the differences among areas in digital IPO rates per million residents can be statistically explained in this framework. (In Equation 3, in other words, the adjusted R2 is .75.)

he implication is that the five areas just mentioned have unusually high rates of IPO activity because large numbers of technically talented people are concentrated in places offering relatively easy access to venture capital—including not only the funding, but the management expertise that comes with it.

Still, given the element of hucksterism that permeated IPOs during the Internet Gold Rush of the late 1990s, it seems advisable to compare the IPO results to more traditional measures.


The obvious question, in other words, is whether we might find a better indicator with which to monitor changes in innovative performance in the U.S. system of cities.

An approach that is sometimes advanced is to compare patent rates in competing metropolitan areas. For example, O‘hUallachain (1999, p. 613) observes, “Innovation is not the product of lone individuals nudging technology forward, but encompasses many interdependent people, firms, and institutions working within networks of social and economic relations.” It turns out, however, that the article is not about innovation at all, but about patents, tabulated relative to population in U.S. metropolitan areas in 1996.

In a similar spirit, Varga observes, “This chapter, using a large data set of US patents, presents the first industrially and spatially detailed analysis of recent trends of innovative activity in the United States” (Varga 1999, p. 230).

Patents Measure Invention, Not Innovation

Unfortunately, and apart from any other limitations of patent data, patents do not measure innovation. Formally, of course, patents are granted by the U.S. Patent Office when it accepts applications to register new ideas, whether for business procedures (as in the recent Amazon single-click case) or for new hardware or industrial processes—or (notoriously, of late) for a chemical formula to be used by pharmaceuticals companies. The patent then confers monopoly rights to the holder, normally for a period of 20 years.

Innovation is a separate step: the commercialization of invention. In a Schumpeterian framework, for example, the four key processes are invention, innovation, emulation, and diffusion. The invention, which may or may not get patented, is the initial idea. The innovation is the process of putting the idea into practice for a profit. Emulation is what happens when competitors “swarm” to provide the same product at a lower price, subject to patent restrictions (as when Compaq reverse-engineered BIOS chip for the IBM PC in the mid-1980s, opening the doors to clones). “Diffusion” refers to the time interval required for an innovation to become widely adopted.

In case the difference is not clear, consider the potentially unnerving case whereby British Telecom is considering pressing a claim that it had applied for (in 1980) and received (in 1989) a patent for a process closely resembling if not identical to the hyperlink. This realization occurred by accident only in the year 2000, when someone stumbled upon an old patent record. (Bray 2000, p. D1.) If such a patent exists and proves valid, the question arises—what happened? A certain rough justice might be served, in that hyperlinks were first joined to the Internet by an Englishman, Tim Berners-Lee, in a project at CERN, the European particle physics consortium in Switzerland, in 1990.

The point here, however, is that the idea was never put into practice by British Telecom, who seem not to have known what to do with it. If so, they had plenty of company in the numerous U.S. managerial corporations who came up with ideas and then had no clue how to proceed. The classic example, among many, is Xerox, whose Palo Alto Research Center (PARC) came up with a dazzling series of revolutionary PC ideas, not one of which Xerox ever commercialized—because they did not improve Xerox’s position in the copier market.

In organizational terms, patents can perhaps best be understood as a running tabulation of what happens in corporate R&D labs, as a look at the U.S. Patent Office’s top 10 patenting organizations will tend to confirm. In 1996, for example, the 10 U.S. organizations with the largest number of patents were IBM, Motorola, the U.S. government, General Electric, Eastman Kodak, Xerox, Texas Instruments, 3M, AT&T, and Hewlett-Packard (O’ hUallachain, p. 624). Only two of these, T.I. and H-P, are from the South or West, and they are both anomalies in their own regions by virtue of their relatively advanced age.

How accurate are patents as indicators of corporate and other invention? A recent study by Cohen, Nelson, and Walsh (2000) points up the indicator’s limitations. Surveying 1478 R&D labs in U.S. manufacturers in 1994, they found that of the several ways firms “protect the profits due to invention…patents tend to be the least emphasized by firms in the majority of manufacturing industries and secrecy and lead time tend to be emphasized most heavily.” By the same token, when patents were employed, it was not necessarily to protect a new discovery but alternatively to attain negotiating leverage, to block other firms’ patents of related discoveries, or to prevent suits.

Even as measures of inventive activity, in short, patents leave something to be desired.

What Do Patents Show about the Geography of R&D or Inventive Activity?

Taken on their own terms, what do such studies of the geography of patent activity reveal? As might be expected, adding location as a dimension creates new measurement issues.

One concerns the location of the discovery itself vs. the location of the patent’s ownership. Tabulations that locate patent activity according to where the patent is owned, not where the inventor (or R&D lab) is located, tend to distort the picture, as when the 23 percent of Arizona’s patent activity attributable to a Motorola facility there in 1996 might have been credited to the parent company’s state, Illinois. Another is that patents have traditionally not covered software code, which instead comes under copyright laws. To that extent, patent counts will tend to slight metros specializing in I.T. (Both these cautionary points are made by O’ hUallachain, p. 628).

Such quibbles aside, what do the two recent patent studies reveal about the U.S. system of cities? Both tend to bring out the “inventiveness” of traditional metros in the Manufacturing Belt. O’ hUallachain, for example, finds that the 87 metros of the Manufacturing Belt accounted for half of all metropolitan patents in 1996, when they had only 44 percent of the metropolitan population. Accordingly, “Metropolitan residents in the manufacturing belt remain the most industrious inventors” (p. 613).

Varga’s findings differ because he monitors changes in patent activity over time, from 1983 to 1992. He finds a general shift in patent activity from the metros of the Manufacturing Belt to areas in the South and West, led by patents registered for I.T. On the other hand, some centers in the Belt retained strong presences in chemicals and pharmaceuticals, and in high-technology machinery. Philadelphia, for example, remained strong in the former, and Chicago in the latter—indeed, Chicago ranked second in 1992 among all areas in terms of high-technology patents. (Varga, p. 225).

The Stellar Patent Performance of the Three Super-States, 1978-1998

Relying purely on patent data, then, the two studies together suggest that the strong performance of Manufacturing Belt metro areas in 1996 may have been a legacy effect. This impression holds up when we perform a new comparison of state patent data over time. We can begin with the 10 states that had most patents in the late 1970s. Seven were from the Manufacturing Belt, and the other three were California, Florida, and Texas, the proverbial “super-states” when it comes to population and employment growth. (The data set has been compiled and provided by Brian Ceh, who also alerted me to the increasing prominence of the latter three states.)

Among the 10 major states in terms of late-1970s patent activity, we can compute the increase in patents generated over the next 20 years. The national count doubled (from 44,762 to 90,676, up 103 percent). But counts roughly tripled in Florida, California, and Texas. As FIGURE 18 shows, Massachusetts came in at the national average, while the remaining six states had increases of less than two-thirds the U.S. pace.

A line is also included in Figure 18 to show state population changes over the same interval. With the possible exceptions of Massachusetts (where patents “outperformed” population, as it were) and New Jersey (where the opposite can be seen), the two indicators show a remarkable correspondence.

The conclusion? For inventive activity no less than for population, the U.S. experienced a pronounced shift away from the Manufacturing Belt in the 1980s and 1990s. On average, in other words, the three “super-states” had increases in patent activity at least triple that of such traditional industrial states as New York, Michigan, Ohio, Illinois, Pennsylvania, and New Jersey.

This brief look at patents, though hardly definitive, suggests four plausible conclusions:

  1. The results reveal that “per capita inventiveness” as a measure is likely to underestimate the speed of the regional transformation, because both patent activity and population have shifted at a rapid pace.
  2. The widely noted U.S. comeback in patent activity during the 1990s (which defied predictions by analysts such as Michael Porter) has depended directly on the supercharged patent performance of the growing states in the South and West, as symbolized here by Florida, Texas, and California.
  3. In the end there is not much difference in the geographical implications as between IPOs and patents as indicators of the geographical dispersal of creativity over the past few decades.  Both indicators, the one of innovation, the other of inventiveness, point up the growing prominence of younger metros and regions within the economy as sources of technological advance.
  4. The greatest exceptions to point (3) are the two resurgent industrial cities, New York (a major IPO seedbed) and Chicago, buoyed by “high-tech” (but not I.T.) patent activity. Here the indicators give different results, and each must be respected.


“Large urban places are not anachronisms in the information age, they are the dominant places in the information age” (Drennan 1999, p. 314).  But which ones, specifically, emerge from among our working list of 30 of the largest U.S. metropolitan areas? In Galbraith’s telling term, which have emerged as strategic cities, places offering “the information that decision requires”?  In light of the indicators we have considered, the first 10 or so areas come quickly to mind, although the exact order remains subjective:

  • San Francisco/Silicon Valley, of course.
  • These days, (2) Boston and Route 128 (again).
  • Washington, D.C., which thanks to northern Virginia is almost as prominent in telecommunications as in government.
  • Dallas, especially when understood as the center of the emerging complex of Texas cities.
  • Seattle, as symbolized by the world’s largest philanthropy, the Bill and Melinda Gates Foundation.
  • Los Angeles, like New York a media-rich location in an age of convergence between content and the Internet.
  • Denver, a relatively unknown powerhouse and a regional capital.
  • New York, by virtue of finance and media.
  • Chicago, another “industrial city” like New York and Boston, and a standout in terms of Old-Economy high-tech patents.
  • Either Atlanta, by the indicators and as a regional capital, or
  • San Diego, an amenity center now liberated from but still technically enriched by its long-time military dependency.

But I see that this top-10 list, loosely based on the empirical indicators we have surveyed, omits such obvious smaller candidates as Albuquerque, Austin, Boise, Miami, Minneapolis, Orlando, Portland (Oregon), or the North Carolina complex—on the basis of size alone. Needless to say, such smaller centers are increasingly prominent in national and global networks of research, production, and innovation.

Come to think of it, recognizing that some other highly innovative cities have been omitted from our list is as good a way as any to do justice to the energizing geography of the New Economy.

Next Section