II. THE ORIGINS OF GEOGRAPHICAL ECONOMICS

There was a long tradition, primarily in Germany from Johann Heinrich von Thünen in Reference Thünen1826 through Wilhelm Launhardt (Reference Launhardt1885) and Georg Pick’s mathematical appendix to Alfred Weber (Reference Weber1909) to Andreas Predöhl (Reference Predöhl1928) and August Lösch (Reference Lösch1940) but also in Sweden with Tord Palander (Reference Palander1935), of using economic theory and formal modeling to find the equilibrium location of economic activities, although using partial equilibrium analysis rather than general equilibrium (Ponsard [1958] Reference Ponsard1983; Blaug Reference Blaug1979). This tradition had indeed received little attention in more general economic theory before World War II, and even less in geography, where search for general explanatory theories was overshadowed by an emphasis on description and areal differentiation associated with Richard Hartshorne ([1939] Reference Hartshorne1961). But from the 1940s onward Walter Isard, originally an economistFootnote ³ (with a Harvard PhD in economics and early publications in the Quarterly Journal of Economics), took the lead in establishing “regional science” as a new discipline between economics and geography, emphasizing neoclassical economic theory of constrained optimization together with formal mathematical and statistical techniques such as interregional input-output models, linear programming, and gravity models, culminating in Location and Space-Economy (Isard Reference Isard1956).Footnote ⁴ Isard (Reference Isard1956, pp. 15, 30–31n, 87n, 93n, 107n, 120–122, 125, 128n, 132, 225n) made frequent reference to an earlier, “highly valuable” Harvard dissertation on The Theory of Geographical Location of Economic Activity by William H. Dean, Jr. (Reference Dean1938), one of the first African-Americans to receive a PhD in economics.Footnote ⁵

Geography itself took a “quantitative turn” in the 1960s, exemplified by Torsten Hägerstrand (Reference Hägerstrand1953, translated from Swedish 1967), William Garrison’s “Spatial Structure of the Economy” (Reference Garrison1959–60), William Bunge’s Theoretical Geography (Reference Bunge1962), Peter Haggett’s Locational Analysis in Human Geography (Reference Haggett, Haggett, Cliff and Frey1965), Richard Chorley and Haggett’s Models in Geography (Reference Chorley and Haggett1967), and David Harvey’s Explanation in Geography (Reference Harvey1969), with a notable forerunner in Edward Ullman (Reference Ullman1941), little-noticed when first published (two years before Ullman completed his PhD) but the only pre-1958 article reprinted in William Leahy, David McKee, and Robert Dean (Reference Leahy, McKee and Dean1970). Ian Burton (Reference Burton1963) wrote of a “quantitative revolution” in geography that was over because its core ideas had been accepted throughout the discipline (see also Bunge Reference Bunge1979; Johnston Reference Johnston1979, ch. 3; Billinge, Gregory, and Martin Reference Billinge, Gregory and Martin1984; Johnston Reference Johnston, Backhouse and Fontaine2010). The central place theorist Brian Berry (Reference Berry1978, pp. vii, ix) held that “by the mid-1960s … [t]he Hartshornian mainstream had withered under theoretical-quantitative attack. To a new generation of young geographers, progress in geography meant theory-building, theory-testing, and theory-refining in an accumulative and unidirectional sequence, and the theory was location theory—von Thünen, Weber, Christaller, and Lösch,” ending the days when “Proponents of alternative paths [were] dismissed as malcontents, kooks, or worse—witness Richard Hartshorne’s dismissal of competing views in The Nature of Geography [(1939) Reference Hartshorne1961] and Perspective on the Nature of Geography [Reference Hartshorne1959].” Several leading figures in the quantitative turn in geography published in Papers and Proceedings in Regional Science (e.g., Berry and Garrison Reference Berry and Garrison1958; Ullman Reference Ullman1962) and held office in the Regional Science Association (Edward Ullman as president in 1960–61, William Garrison in 1962–63, and Torsten Hägerstrand in 1968–69, and Brian Berry as one of two vice-presidents in 1966–67), while the election of Walter Christaller as a vice-president for 1961–62 provided a symbolic link to the earlier German tradition of location theory. The titles of Bertil Ohlin’s International and Interregional Trade (Reference Ohlin1933) and Isard’s “Location Theory and Trade Theory” (Reference Isard1954) show international trade theorists and regional scientists before Krugman recognizing the analogy between international trade and the location of economic activities, with countries as just one kind of region. However, the perfect competition and constant returns assumption then prevalent in economic theory did not suffice for general equilibrium analysis of location decisions, so location theory remained at the periphery of mainstream economics, taught, if at all, in field courses on urban and regional economics. What was missing, both in international trade theory and in location theory in regional science and geography, was a tractable general equilibrium analysis incorporating increasing returns to scale.

The idea of increasing returns to scale goes back to Adam Smith’s dictum that “the division of labour is limited by the extent of the market” (Smith [1776] Reference Smith and Skinner1970, Bk. I, ch. 3) and to Alfred Marshall’s external economies in the 1890s but long resisted tractable formalization. Avinash Dixit and Joseph Stiglitz’s “Monopolistic Competition and Optimum Product Diversity” (Reference Dixit and Stiglitz1977) combined monopolistic competition (pioneered by Edward Chamberlin Reference Chamberlin1933) with increasing returns to scale to create a tractableFootnote ⁶ model of what products will be produced, a model in which increasing returns produced clustering in a general equilibrium framework without going to an equilibrium of complete monopoly. Paul Krugman’s Nobel Prizewinning contribution to international trade theory (Krugman Reference Krugman1980, Reference Krugman1990) transferred this model from product space to actual space, the location of economic activities, with agglomeration resulting from economies in transport costs.Footnote ⁷ Krugman (Reference Krugman1991a, Reference Krugman1991b) then extended the Dixit-Stiglitz model from international trade to economic geography and combined it with the assumption made by Paul Samuelson (Reference Samuelson1952) that transport cost can be viewed as the transported good melting like an iceberg at a steady rate over time, presenting this mathematical modeling of general equilibrium in space as a formal, rigorous replacement for the supposedly descriptive, untheoretical work of geographers. As Richard Baldwin et al. (Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003, pp. 227–228, 230) emphasize, the resulting “models have a number of properties that make policy analysis very different to standard neoclassical models … because of the non-linearity of the models, which produce multiple equilibria, hysteresis and bifurcations,” with such consequences as making “small policy interventions ineffective when it comes to location” until a threshold is passed and “a massive dislocation of industry” occurs, and that, because of hysteresis, “[b]ad policies, even when they are temporary, may have long-lasting bad effects.” In the new models, “many policies have very non-linear and even non-monotonic effects on prices, trade flows, industrial location, and the like. Moreover, the impacts of policies often interact with trade openness, sometimes in unexpected ways” (Baldwin et al. Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003, p. 472).

Even the crucial modeling strategy of Krugman (Reference Krugman1991a, Reference Krugman1991b), the adaptation of the Dixit-Stiglitz general equilibrium model of increasing returns and monopolistic competition to the explanation of agglomeration in economic geography, had, unknown to Krugman,Footnote ⁸ been undertaken by regional scientists publishing in Regional Science and Urban Economics (a journal that bridged the separated discipline of regional science and a subfield of economics, urban economics, but was unlikely to be read by an international trade theorist), Masahisa Fujita (Reference Fujita1988) and his doctoral student Hesham Abdel-Rahman (Reference Abdel-Rahman1988) (see also Fujita Reference Fujita1989; Abdel-Rahman and Fujita Reference Abdel-Rahman and Fujita1990; Abdel-Rahman and Wang Reference Abdel-Rahman and Wang1995; Abdel-Rahman Reference Abdel-Rahman1996). Fujita took an undergraduate degree in urban planning from Kyoto University before doing a PhD with Walter Isard in the Department of Regional Science that Isard had founded at the University of Pennsylvania, then became a professor in that department where he supervised Abdel-Rahman’s 1987 dissertation. The University of Pennsylvania eventually abolished its Department of Regional Science after thirty-five years in 1993Footnote ⁹ (see Isserman Reference Isserman1993; Barnes Reference Barnes2003, Reference Barnes2004a), and Fujita returned to Kyoto as a professor of economics.

The much greater attention paid to Krugman than to Fujita and Abdel-Rahman was not due to any difference in policy relevance, which all three authors stressed, or to Krugman addressing questions rather than just formal modeling, since all three did so (see, e.g., the title of Abdel-Rahman Reference Abdel-Rahman1996). All three asked policy-relevant questions about how agglomeration economies cause economic activities to be concentrated in some places rather than others. Krugman (Reference Krugman1991b, pp. 2–3) gave the explanation when, after regretting that “nobody is … busy exploiting the facts and insights that can come from looking at localization and trade within countries,” he qualified this “unfair statement” by noting, “There are excellent economic geographers out there, as well as urban and regional economists who worry about geographical issues … however, these people are almost uniformly peripheral to the economics profession” whereas “International economics is a flagship field.” Economic geographers and urban and regional economists “may do excellent work, but it does not inform or influence the economics profession” unless and until independently reinvented by people in a flagship field. The circularity of the argument, that international trade theorists need not heed the work of economic geographers or urban and regional economists (“nobody”) because their work is not heeded by international trade theorists, does not diminish its force as a description of how economics is conducted (Krugman Reference Krugman1991b, p. 3, remarked, “There are good reasons why this has happened and equally good reasons why it should change”).

Priority issues were quietly settled by some joint publications starting with Fujita and Krugman (Reference Fujita and Krugman1995 in Regional Science and Urban Economics; see also Fujita and Krugman Reference Fujita, Krugman, Florax and Plane2004) and co-authorship of a major treatise and textbook on The Spatial Economy (Fujita, Krugman, and Venables Reference Fujita, Krugman and Venables1999) but, although The Spatial Economy was widely cited, Krugman remained very sparing in mention of Fujita’s earlier work (and Abdel-Rahman’s name is unmentioned in The Spatial Economy or Krugman Reference Krugman1995).Footnote ¹⁰ The only mention of Fujita in Krugman’s Ohlin Lectures Development, Geography, and Economic Theory (Reference Fujita and Krugman1995, p. 90) is a passing remark in the appendix of:

Fujita (Reference Fujita1988) appears in the bibliography of Fujita, Krugman, and Venables (Reference Fujita and Krugman1995), but no index entry leads the reader to any mention of that article, let alone a statement that it made use of Dixit and Stiglitz (Reference Dixit and Stiglitz1977). Fujita was mentioned in Krugman (Reference Krugman1998a) only for works co-authored with Krugman (see Rosser Reference Rosser2011).

The most notable treatment by an historian of economic thought,Footnote ¹¹ Stephen Meardon (Reference Meardon, Backhouse and Biddle2000), of the emergence of geographical economics gave a good account of the relationship of the “new economic geography” to the pre-World War II tradition of German location theory but did not mention Fujita, even as a co-author of Krugman. Meardon (Reference Meardon, Backhouse and Biddle2000, p. 346) drew attention to J. Vernon Henderson’s comment on Krugman (Reference Krugman1996), where “Henderson [Reference Henderson1996] reminded Krugman that neoclassical urban systems models incorporated imperfect competition, in some cases the very same Dixit-Stiglitz variety, even before Krugman (Reference Krugman1991[a]) launched the new economic geography.” Meardon (Reference Meardon, Backhouse and Biddle2000, pp. 355–356) quoted the condescending evaluation by Krugman (Reference Krugman1995, p. 87) that “One cannot fault the geographers for their failure to develop full maximization and equilibrium models although one can perhaps complain about their failure to understand how short of that ideal they were falling” but added that geographers such as Ron Martin (Reference Martin1999) disputed Krugman’s narrative.

Apart from Meardon (Reference Meardon, Backhouse and Biddle2000), the relevant history of thought literature is thin, with the promising exception of a bibliometric study by Anthony Rebours (Reference Rebours2018) of the interactions between geography and economics in recent decades. Jacques-François Thisse (Reference Thisse2011, Reference Thisse, Faccarello and Kurz2016) provides a practitioner’s history of the emergence of geographical economics, discussing von Thünen (Reference Thünen1826–63) and Harold Hotelling (Reference Hotelling1929) as a backdrop to celebrating Krugman (on spatial competition models since Hotelling, see Eaton and Lipsey Reference Eaton and Lipsey1997, collecting their earlier articles). Although himself a co-author of Fujita since Reference Fujita and Thisse1986, Thisse (Reference Thisse, Faccarello and Kurz2016) is reticent about Fujita’s early work, citing his book Urban Economic Theory (Fujita Reference Fujita1989) but not Fujita (Reference Fujita1988) or anything by Abdel-Rahman. In his successful textbook on the history of geographic ideas, Geoffrey Martin (Reference Martin2005, p. 509) offers only a puzzling remark about Krugman “embracing the notion that comparative cost advantages may be ignored given specialization and trade that secure increased profit” (an interpretation that he attributes to Trevor Barnes) and the horrified (but undeniable) observation that “Krugman plans to make economic geography a branch of economics (Krugman Reference Krugman1995).” Sixty years after his doctorate, and more than sixty years after he began publishing in journals, Walter Isard published his History of Regional Science and the Regional Science Association International (Reference Isard2003), but, padded with the programs of conferences and lists of council members, this book (like Isard Reference Isard1979) lacked the force of his earlier works. Of thirty-five items in the bibliography, sixteen were by Isard and the others included Bertil Ohlin (Reference Ohlin1933), Samuelson’s introductory textbook, Kenneth Arrow and Gerard Debreu on the existence of general equilibrium, John Maynard Keynes, Walt Rostow, and Jan Tinbergen. Trevor Barnes (Reference Barnes2004b, p. 222–223) dismissed Isard (Reference Isard2003) as “really a scrapbook of the history of regional science, with almost no interpretation other than an underlying triumphalism.… Missing from this volume … is an explanation of [regional science’s] success, or any recognition of its more recent failure.”

III. TWO APPROACHES TO ECONOMIC GEOGRAPHY

The geographical economics presented in the Journal of Economic Geography, the Oxford Handbook of Economic Geography (Clark, Gertler, and Feldmann Reference Clark, Gertler and Feldman2000), in Fujita, Krugman, and Venables (Reference Fujita1999), or in the treatise and textbook Economic Geography and Public Policy by Baldwin et al. (Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003) (notwithstanding some notable attempts at communication “from the other shore” by Scott Reference Scott2000; Sheppard Reference Sheppard, Clark, Gertler and Feldman2000, Reference Sheppard2001; Barnes Reference Barnes2004a, and notwithstanding that the Oxford Handbook has mirror chapters by an economist and a geographer and that the editorial board of the Journal of Economic Geography is balanced between economists and geographers) can be sharply distinguished by inspection from the economic geography typical of geography departments, exemplified by the journal Economic Geography; A Companion to Economic Geography (Sheppard and Barnes Reference Sheppard, Clark, Gertler and Feldman2000); or Allen Scott’s Clarendon Lectures, Geography and the Economy (Reference Scott2006), or his Regions and the World Economy (Reference Scott1998). This contrast may seem paradoxical, given that the radical geographers Barnes, Scott, and Eric Sheppard served on the editorial boards of both journals and contributed to the Oxford Handbook while, in the case of Sheppard and Barnes, editing the Blackwell Companion. But the difference is clear: contributions by economists (and some geographers) in the Oxford Handbook or the Journal of Economic Geography are unmistakably mainstream neoclassical economics (e.g., Krugman Reference Krugman, Clark, Gertler and Feldman2000), and the engagement and dialogue promoted there between geography and economics are between geography and neoclassical economics (geographical economics), leaving David Ricardo, Karl Marx, and Piero Sraffa outside modern economics but perhaps in geography (all three are in the subtitle of Sheppard and Barnes Reference Sheppard and Barnes1990). In contrast, in the Blackwell Companion, neoclassical economics, far from being accepted as the imperial social science, was covered by a single chapter (by a geography lecturer) on “The Modeling Tradition” on an equal footing with chapters on Marxian, feminist, institutionalist, and post-structuralist approaches,Footnote ¹² while readers of Economic Geography might find a six-paper symposium on the engagement between geography and a variety of heterodox economics, evolutionary political economy (in the April 2009 issue).

The difference between the two approaches to economic geography is shown by what Fujita, Krugman, and Venables (Reference Fujita1999, p. 45) write about their “workhorse model”: “Dixit-Stiglitz monopolistic competition is grossly unrealistic, but it is tractable and flexible; as we will see, it leads to a very special but very suggestive set of results.… This spatial Dixit-Stiglitz model is a crucial ingredient in almost everything that follows.” This core-periphery model, described by Baldwin et al. (Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003, p. 2) as having “the unfortunate feature of being astoundingly difficult to work with analytically,” often requiring approximating the multiple equilibria numerically with a search algorithm rather than finding them by an analytical solution, was made more tractable by Baldwin et al. (Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003) and Rikard Forslid and Gianmarco Ottaviano (“An Analytically Solvable Core-Periphery Model,” Reference Forslid and Ottaviano2003).Footnote ¹³ The point is not the specific model, or the use of formal quantitative methods or explicit theorizing, but the willingness to commit to a “grossly unrealistic,” “astoundingly difficult,” highly mathematical modeling strategy provided only that it yields “very suggestive results” within the general equilibrium framework standard among economists.

Another striking difference is that, after its “quantitative turn” in the 1960s, a substantial part of human geography took a different turn from the 1970s onwards, indicated by the title question of Ash Amin and Nigel Thrift (Reference Amin and Thrift2000), “What Kind of Economic Theory for What Kind of Economic Geography?”; by the title of New Models in Geography (Peet and Thrift Reference Peet and Thrift1988), named to challenge Models in Geography (Chorley and Haggett Reference Chorley and Haggett1967); and by the subtitle of Sheppard and Barnes (Reference Sheppard and Barnes1990), Geographical Analysis After Ricardo, Marx and Sraffa. Allen Scott, once a faculty member in regional science at the University of Pennsylvania (later Distinguished Professor of Public Policy and Geography at UCLA) and joint secretary with David Harvey of the British Section of Isard’s Regional Science Association, also discovered Marx and Sraffa (Barnes Reference Barnes2004a, p. 124). David Harvey, once the author of Explanation in Geography (Reference Harvey1969) and later Halford Mackinder Professor of Geography at Oxford, wrote a companion to each of the three volumes of Karl Marx’s Capital, as well as Seventeen Contradictions and the End of Capitalism (Harvey Reference Harvey2014),Footnote ¹⁴ the dust jacket of which reports, “He is among the twenty most-cited authors in the social sciences and humanities and is the world’s most cited academic geographer” (although by then he had crossed disciplinary lines as Distinguished Professor of Anthropology at the Graduate Center of the City University of New York). Barnes, Harvey, Scott, Sheppard, and other defectors from Isard’s regional science did not reject economic theory, but the Marxist or Sraffian economic theory that they embraced was not at all the neoclassical, general equilibrium economic theory of Krugman and Fujita. This sharp contrast in economic methodology does not, of course, contradict the position of Krugman (or, for that matter, Stiglitz) on the more left end of the US political spectrum, as is evident to any reader of Krugman’s op-ed column in the New York Times.

What the radical geographers, whether Marxist or Sraffian (neo-Ricardian), took from David Ricardo’s Principles ([1817] Reference Ricardo, Sraffa and Dobb1951) was not what Krugman or other trade theorists or classical and neoclassical location theorists took from Ricardo. Krugman (Reference Krugman1990, p. 4) wrote, “The long dominance of Ricardo over Smith—of comparative advantage over increasing returns—was largely due to the belief that the alternative was necessarily a mess” until Dixit and Stiglitz (Reference Dixit and Stiglitz1977) managed to formalize increasing returns in a tractable model. While trade theorists emphasized the numerical example of comparative advantage in Ricardo’s Chapter 7, “On Foreign Trade,” with Portugal specializing in wine and England in cloth even though Portugal has an absolute advantage in both goods, location theorists looked to the classical theory of rent, which von Thünen extended to differences in location of land from Ricardo’s margin of cultivation among plots of land differing in fertility. The radical geographers followed Marx and Sraffa in looking to Ricardo’s theory of value, not his analysis of gains from trade or of the margin of cultivation. This variety of messages to take from Ricardo has made Piero Sraffa and Maurice Dobb, the editors of his works, the only Communists reprinted by the Liberty Fund (Ricardo [1817] Reference Ricardo, Sraffa and Dobb1951).

IV. FIVE LOOPS OF THE “NEW ECONOMIC GEOGRAPHY”: AN INTERPRETATION AFTER JOHNSON 1971

Harry Johnson held in his Ely Lecture, “The Keynesian Revolution and the Monetarist Counter-Revolution” (Reference Johnson1971), that the success of a revolutionary approach to economics depends on five core characteristics. First, the new idea must confront the principal hypothesis of the conservative orthodoxy and reverse that main hypothesis in an academically acceptable way. Second, it must give the impression that it is a new theory yet retain, under new names, the effective factors or unquestionable elements of the prevailing orthodoxy. Third, the new theory must also be sufficiently difficult to understand so that established senior academic individuals will not master its technical aspects, providing opportunity for younger academics to advance their careers by doing so. It must be difficult enough to test the talent of the younger scholars but easy enough so that they can successfully understand it. Fourth, the new theory has to provide scholars with a new attractive methodology as compared with the pre-existing ones, which will consist of an extensive mathematical approach. This new methodology will require the scholars to have a broad empirical knowledge base to understand it. Last, empirical relationships are introduced for the econometricians to estimate the relationship between components.

These five factors made the Keynesian revolution successful, displacing the pre-existing orthodoxy to become an orthodoxy itself. However, the Keynesian revolution was a victim of two characteristics: inability to cope with a major social problem (inflation), and dependence on the authority and prestige of senior scholars whose position barred the advancement of younger academics. Thus, the Keynesian revolution was as vulnerable to revolution as the previous orthodoxy. The monetarist counter-revolution became successful as it provided the explanation to a policy-relevant problem that the Keynesians could not explain and therefore was received as the new “new economics” in place of the Keynesian “new economics.” Thus, according to Johnson (Reference Johnson1971), first Keynesianism and then monetarism revolutionized macroeconomics by giving new confusing names to parts of the existing orthodoxy, changing its methodology by adding mathematics and empirical relationships, and providing opportunity for a new generation of scholars to advance their careers.

First, the new idea will have to confront the principal hypothesis of the conservative orthodoxy and it must be academically acceptable, which can reverse the main hypothesis. It means that the new idea will have to be similar to an existing orthodoxy but will have to offer liberation from the established primary orthodoxy. Contrary to the established belief in economics and geography that increasing returns to scale could not explain cities because increasing returns models would imply a single huge city, Krugman (Reference Krugman1991a, Reference Krugman1991b) and, independently of him, Fujita (Reference Fujita1988) and Abdel-Rahman (Reference Abdel-Rahman1988) showed that the Dixit-Stiglitz combination of monopolistic competition with increasing returns could yield an equilibrium distribution of agglomeration within a general equilibrium framework. These theoretical developments were presented as having policy relevance, like the changes in macroeconomics considered by Johnson. Abdel-Rahman (Reference Abdel-Rahman1988) linked his use of the Dixit-Stiglitz model for characterizing equilibrium and optimum city size to a subsidy scheme to achieve the first-best optimum city size. Abdel-Rahman and Fujita (Reference Abdel-Rahman and Fujita1990, p. 165) invoked “normative or policy questions” as the motivation to “develop an alternative framework which models agglomeration economies endogenously, explicitly showing their origins.”

Second, the new theory must give the impression that it is a new theory but still consist of the effective factors or at least not readily disputable components of the prevailing orthodoxy. Like the “regional science” founded by Walter Isard (Reference Isard1949, Reference Isard1956) but without much awareness of regional science or of the quantitative turn in geography, Krugman’s “new economic geography” provided, as an alternative to economic geography (as it was before the quantitative turn) a rigorous analytical framework derived from mathematical models. What made Krugman’s “new economic geography” new was the use of the Dixit-Stiglitz model of monopolistic competition and increasing returns, an innovation that was also introduced by the regional scientists Fujita (Reference Fujita1988) and Abdel-Rahman (Reference Abdel-Rahman1988), trained in Isard’s Department of Regional Science and steeped in the long tradition of location theory, as indicated by Fujita’s writings on von Thünen, Hotelling, and Isard.

Third, the new theory must also incorporate a high, but not too high, level of difficulty to apprehend. It will have to be so difficult that senior academic authors will not be willing to master the theory and be difficult enough to test the talent of the younger students but also easy enough so that the younger scholars can understand it with sufficient intellectual effort. The NEG is modeled using rigorous mathematical causal models, which are used to evaluate empirical analysis. The first generation of NEG models relied heavily on rather specific modeling assumptions or tricks (for example, Krugman and Fujita were both inspired by the Dixit-Stiglitz model of monopolistic competition, which relied heavily on numerical models as well). This shows that NEG was dressed up in mathematics that excluded the work of economic geographers as the new model was too mathematical for the previous generation to apprehend. According to Barnes (Reference Barnes2004a), the economic geographer would require a massive retro-education in formal theory and techniques if they want to understand and participate in NEG. This proves the first half of the third characteristic of Johnson. Similarly, simplification was made for alternative utility and demand functions and production structures (Baldwin et al. Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003; Forslid and Ottaviano Reference Forslid and Ottaviano2003) that made the second generation of NEG models analytically solvable and easier to handle by the younger academic individuals when intellectual efforts were made.

Fourth, the new theory has to provide the gifted and less opportunistic scholars with a new attractive methodology compared with the pre-existing ones. This new methodology will require the scholars to have a broad mathematical as well as empirical base to understand it. As indicated above, NEG was designed in a general equilibrium model of monopolistic competition. Fujita, Krugman, and Venables (Reference Fujita1999) state that there are three classes of models: regional models, urban system models, and international models. The regional model basically includes the core-periphery model, introduced in Krugman (Reference Krugman1991a), as a basic introductory framework for the new economic geography—a framework that illustrates how the interactions among increasing returns at the level of the firm, transport costs, and factor mobility can cause spatial economic structure to emerge and change. On the other hand, according to Harry Garretsen and Ron Martin (Reference Garretsen and Martin2010), economic geography by geographers is highly focused on cultural, social, and institutional accounts, which were disregarded by NEG theorists mainly because geographers did not focus narrowly on the formal modeling and hypothesis testing with which economists are comfortable. Hence, this shows that the methodology provided by “new” theory is more methodologically attractive to economists compared with the ones provided by the geographers, and hereby this meets the fourth characteristic of Harry Johnson.

Last, the fifth characteristic emphasizes the empirical relationship introduced by the econometricians to estimate the relationship between components (see the survey by Head and Mayer Reference Head, Mayer, Henderson and Thisse2004). The NEG provides ample opportunity for the sort of empirical work for which applied economists are trained and proficient. The model has examined the relationships between market access and wages and between market access and industrial location, and has confirmed the central characteristic of economic geography, which is the existence of multiple equilibria (Redding Reference Redding2008). This shows that NEG not only provides a theoretical base but also emphasizes study of empirical relationships, which provides econometricians and applied economists an opportunity to measure relationships between specific components in ways not previously done by geographers.

V. THE GERMAN TRADITION OF LOCATION THEORY

The work of the German location theorists played a major role for the foundation of regional science. Isard (Reference Isard1949) cited, as influential for the emergence of regional science, in chronological order of relevant publication: Johann Heinrich von Thünen (Reference Thünen1826–63), Albert Schäffle, Wilhelm Roscher, Wilhelm Launhardt (Reference Launhardt1885), Alfred Weber (Reference Weber1909), Andreas Predöhl (Reference Predöhl1928), Oskar Engländer, Hans Ritschl, Hans Weigmann, Walter Christaller (Reference Christaller1933), Erich Schneider, and August Lösch (Reference Lösch1940) (see Ponsard [1958] Reference Ponsard1983; Blaug Reference Blaug1979). This rich tradition was long neglected by mainstream geography. In his very influential work The Nature of Geography (see also Hartshorne Reference Hartshorne1959; Entrikin and Brunn Reference Entrikin and Brunn1989 for its influence), Richard Hartshorne ([1939] Reference Hartshorne1961, pp. 420–421) barely mentioned location theorists such as von Thünen and Weber, remarking, “American economists in particular have shown little interest in this field and at the time the writer was concerned with it were hardly aware of Weber’s work, which, in any case, Palander has shown to be impracticable.” Although Hartshorne (Reference Hartshorne1927) had once engaged in location analysis, Hartshorne ([1939] Reference Hartshorne1961) emphasized areal differentiation, not the search for explanatory theory. Similarly, on the eve of the quantitative turn in geography, Thomas Walter Freeman’s A Hundred Years of Geography (Reference Freeman1961, pp. 162–163) discussed Lösch on the economics of location only briefly, as an unimportant special topic. Hartshorne ([1939] Reference Hartshorne1961, pp. 99–100) and Freeman (Reference Freeman1961, pp. 149–150) noted, but with little interest, the economist Josiah Stamp’s “Geography and Economic Theory” (Reference Stamp1937), with Freeman remarking that “there are endless complexities in the location of industry” and Hartshorne confusing Josiah Stamp with the geographer L. Dudley Stamp. In contrast to pre-1960s geographers (and, largely, pre-1960s economists), regional scientists such as Isard and Fujita paid close attention to the pre-World War II tradition of location theory.

Von Thünen’s contribution to spatial economics was discussed in the first volume of The Isolated State (Reference Thünen1826–63),Footnote ¹⁵ where he extended the classical theory of rent from differences in fertility of land to differences in location. This volume of von Thünen’s writing was not available in English translation until 1966. In The Isolated State von Thünen assumed a homogenous or isolated space around a city where goods are bought and sold, with production of diverse products organized in rings around the city. This "monocentric city model" provided the theoretical foundation for urban economics. Transport cost determined which commodities were produced in the inner or outer ring. Perishable commodities had high transportation costs, so they secured the inner ring, and the commodities having lower transportation cost secured the outer rings. The area beyond the outermost ring was uncultivated wilderness. The rent of land varied according to proximity to the center: highest for land in the center and zero at the boundary of uncultivated wilderness.

Masahisa Fujita (Reference Fujita2011) termed von Thünen “the founding god” of modern economic geography, including not only traditional economic geography and location theory but also the modern urban economics as well as the so-called new economic geography.Footnote ¹⁶ Location theorists and economic geographers (both “traditional” and “new”) always referred to von Thünen on location of agricultural activity but not in the context of agglomeration economies or city formation, where Fujita and Jacques-François Thisse ([2002] Reference Fujita and Thisse2013) held that von Thünen’s agglomeration factors matched Krugman’s explanation for the emergence of a core-periphery structure. The recognition of von Thünen’s contribution in their book harked to their first joint article, “Spatial Competition with a Land Market: Hotelling and Von Thünen Unified” (Reference Fujita and Thisse1986). Krugman’s stress on novelty together with Fujita’s emphasis on how much was retained from von Thünen, Hotelling, and Isard satisfied Harry Johnson’s criterion that the new approach must give the impression that it is a new theory but still encompass the aspects of the prevailing orthodoxy of continuing validity or usefulness.

Paul Krugman (Reference Krugman1991b, pp 14–15) states:

Von Thünen’s centripetal forces as identified by Johannes Bröcker (Reference Bröcker, Nijkamp, Rose and Kourtit2015)—scale economies due to indivisibilities and to division of labor, economies of shopping, labor market matching in cities, risk sharing—correspond with Krugman’s description of the core-periphery model of the NEG. Fujita explains that, despite the similarities, Krugman clearly highlighted the circular causality between agglomeration of industries and workers using demand externalities whereas von Thünen was less explicit about circular causality. According to Fujita (Reference Fujita2011), recent technological breakthroughs have made it possible for economic geographers to analyze von Thünen’s work better, even though not all his work has yet been brought to light.

Mark Blaug (Reference Blaug1979), in “The German Hegemony of Location Theory: A Puzzle in the History of Economic Thought,” observed that spatial economics has a remarkable history that began with seminal writings by von Thünen and included Alfred Weber’s revolutionary The Location of Industries. Part I. Pure Theory of Location (Reference Weber1909), which extended von Thünen’s analysis of the spatial distribution of agricultural activities to explain the location of industrial plants. Despite these early developments, locational problems received little attention throughout the nineteenth century and first quarter of the twentieth century, turning economics into what Isard (Reference Isard1956, pp. 25–26) termed “a wonderland of no spatial dimensions.”

The model of the location triangle developed by Alfred Weber (elder brother of the sociologist Max Weber) was solved not by Weber himself but in an appendix by George Pick, a mathematics professor in Prague, where Weber had previously held the chair of economics. Weber’s model resembled the model formulated by Carl Wilhelm Friedrich Launhardt in 1882 and 1885 where, building on von Thünen in a more rigorous mathematical style, he illustrates a geometrical solution for three points with weights. Carl J. Friedrich, when translating Weber (Reference Weber1909) into English in 1929, observed that Weber was not acquainted with Launhardt, whose influence is apparent in Pick’s mathematical appendix rather than in the text itself. However, neither Weber nor Pick cited Launhardt (Reference Launhardt1885), and they were reproved accordingly by Ladislaus von Bortkiewicz in a review of Weber’s book (see notes by C. J. Friedrich, the translator, in Weber Reference Weber1929, pp. 238–239; Perreur Reference Perreur1998; Bröcker Reference Bröcker, Nijkamp, Rose and Kourtit2015).

Walter Christaller introduced central-place theory in his book Central Places in Southern Germany (Reference Chamberlin1933), establishing a powerful theoretical base for the study of the factors of location, size, and number, and the geometrical arrangement of cities in space. According to central-place theory the principal purpose of a settlement is the provision of goods and services for the adjacent market area. Lösch’s Economics of Location came out in 1940 as Die Räumliche Ordnung der Wirtschaft (Lösch Reference Lösch1940) and in Reference Stolper and Lösch1954 in English translation by William H. Woglom and Wolfgang F. Stolper (best known for the Stolper-Samuelson theorem in international trade theory). The economist Lösch built his theory based on the work of the geographer Walter Christaller (Reference Christaller1933), taking a deductive approach of formal modeling where Christaller’s book had been more descriptive. In contrast to Christaller, who started with the highest order when dealing with the central place system, Lösch derived partial equilibria for each network of market areas grounded on Edward Chamberlin’s monopolistic competition, finding hexagonal market areas, but he did not take into consideration the problem arising from extending Chamberlin’s symmetry assumption to the spatial market (Beckmann and Thisse Reference Beckmann, Thisse and Nijkamp1986; Bröcker Reference Bröcker, Nijkamp, Rose and Kourtit2015).

VI. MASAHISA FUJITA ON LOCATION THEORISTS FROM VON THÜNEN TO ISARD

Fujita (Reference Fujita, Krugman and Venables1999), in his speech entitled “Location and Space-Economy at Half a Century,” discussed Walter Isard’s major work, Location and Space-Economy (Reference Isard1956), stressing the three basic ideas discussed by Isard on the general theory of location and space-economy. First, the general theory of location and space-economy must be something more than the traditional general equilibrium analysis based on perfect competition. The traditional general equilibrium analysis of Léon Walras, Vilfredo Pareto, and John Hicks did not incorporate spatial dimensions clearly or include the consequence of transport and spatial cost on the economic activities in space, so it could not explain endogenous concentration of economic agglomeration.

Second, general theory should be identical with the general theory of monopolistic competition. Isard referred enthusiastically to the German location theorist August Lösch (Reference Lösch1940) but also to Chamberlin’s seminal work The Theory of Monopolistic Competition (Reference Chamberlin1933), which emphasized product differentiation.

Third, the evolutionary approach should be used to represent dynamic relations in the general theory. The spatial economy is explained using two opposite forces of the centripetal drive of increasing returns and the centrifugal effects of diminishing returns.

VII. CONCLUSION: ECONOMIC GEOGRAPHY, NEW ECONOMIC GEOGRAPHY, GEOGRAPHICAL ECONOMISTS

The essential question in economic geography relates to the explanation of uneven spatial development. Geographers long approached this question descriptively, stressing areal differentiation rather than seeking a generally applicable theoretical framework. From the mid-1950s to the 1970s, geographers increasingly based their research on the ideas of Lösch and other German location-theorists, by means of statistical models and quantitative methods. Many geographers, including several who had been prominent in regional science and in the quantitative turn in geography, subsequently inclined towards Marxist political economy and methods of historical materialism in the 1970s and1980s (Sheppard and Barnes Reference Sheppard and Barnes1990; Amin and Thrift Reference Amin and Thrift2000, Martin and Sunley Reference Martin and Sunley2007; Garretsen and Martin Reference Garretsen and Martin2010; Harvey Reference Harvey2014).

Space and geography did not receive nearly as much consideration from the majority of economists before 1990. Over the past two decades, neglect of geography by economists has been mended by a subfield that is now widely recognized as “new economic geography” (NEG or geographical economics). Krugman (Reference Krugman1991a, Reference Krugman1991b) introduced this approach without knowledge of Fujita (Reference Fujita1988) or Abdel-Rahman (Reference Abdel-Rahman1988) and with minimal attention to Isard, whose Location and Space-Economy (Reference Isard1956) appeared in the bibliography of Krugman (Reference Krugman1991a) but not the index. Fujita and Abdel-Rahman, starting from Isard’s regional science, used the Dixit-Stiglitz model of monopolistic competition to study agglomeration in a general equilibrium framework, emphasizing (as did Krugman) the policy implications of such analysis. All these authors were concerned with understanding uneven spatial economic development, and particularly the phenomenon of spatial agglomeration of economic activity, with their work coming together in Fujita, Krugman, and Venables (Reference Fujita1999). Fujita and Thisse ([2002] Reference Fujita and Thisse2013), Baldwin et al. (Reference Baldwin, Forslid, Martin, Ottaviano and Robert-Nicoud2003), and Proost and Thisse (Reference Proost and Thisse2019) survey the subsequent development of NEG as a systematic and consistent analysis of the location of economic activities.

Geographical economics or the new economic geography emerged from two sources, international trade theory within economics for Krugman and regional science for Fujita and Abdel-Rahman, as scholars from each field extended the Dixit-Stiglitz monopolistic competition model of product diversity to location in space. The emergence of this approach involved, and continues to involve, contestation of disciplinary boundaries among economics, geography, and regional science with the disputes between geographical economists and radical economic geographers being not over whether to use economic theory but about which sort of economic theory to rely upon for spatial analysis.

Krugman, at least until Reference Krugman1995, considered himself to be the sole founder of NEG and held that mathematical models had not been sufficiently utilized by conventional economic geographers while they were dealing with the problems of regional economics. This paper suggests that Harry Johnson’s five characteristics provide a framework to consider how dressing up an existing theory with a new confusing name and attractive mathematical modeling, together with a policy-relevant issue not adequately dealt with in existing theory, could have accounted for the acceptance for NEG. Concepts borrowed from the Dixit-Stiglitz monopolistic competition model in industrial organization assisted Krugman and Fujita in modeling NEG and, in Krugman’s case, the new trade theory, while Isard’s regional science and the long tradition of German location theory shaped the work of Fujita and Abdel-Rahman.

The new economic geography has contributed towards an enormous amount of high-quality research in this field. Notwithstanding the criticisms made towards NEG, it has some characteristics that cannot be neglected, like formal modeling, which helps to detect assumptions giving specific results and maintaining consistency as well as meticulous empirical analysis focusing on categorizing and testing uncertain theoretical predictions. As Krugman (Reference Krugman1998b, p. 7) observed, a theory must last or be superfluous, based on its empirical relevance and success.