I was born in Istanbul to Greek parents. After an education at Robert Academy, the boys’ American high school on the Bosporus, I was offered a scholarship by Carnegie-Mellon University (CMU). I was an intellectual hybrid, trying to balance assertive right and left cerebral hemispheres; one side engaging in drawing and painting and amateurishly trying creative writing; the other, developing a strong interest in geometry, applied mathematics and science. This drama did not end at CMU. I started out in engineering, then transferred to architecture, finally finding a “home” in civil engineering. I wanted to invent a science that would enable the design of a city as a well-functioning social and physical system. But I was quickly humbled by realizing that while cities should be designed for people, unpredictable human behavior necessitated new designs and so on. This led me to economics, a technical enough science of human behavior that made both prediction and prescription possible. At the time, CMU started double degree programs. I graduated with a B.S. in civil engineering and a B.A. in economics. I had also concentrated my electives, putting me a handful of courses short of a degree in philosophy.
Graduate studies at Penn were ideal. Penn allowed an interdisciplinary approach to urban and regional studies with regional science at the center. I was most strongly influenced by Britton Harris. At the end of my first year, I daringly turned down a fellowship offered by Walter Isard, to continue as a research assistant for Britt. Meanwhile, Colin Gannon taught an influential introduction to urban economics at a time when the field was receiving attention from Robert Solow, Avinash Dixit and Edwin Mills (then at nearby Princeton). Alan Wilson and David Pines who visited Penn were also influential.
The monocentric city was the jewel of urban economics. As a student in Colin’s class, I was frustrated by its assumptions: a static model, one dimensional and assuming that all jobs are pinned in the city center. The beauty of economics shone through the model, but the anachronism of the model was disheartening. This tension explains why I took an iconoclastic approach to urban economics. In my dissertation at Penn, I modeled a monocentric city as growing in successive rings, like a tree does, the width of the rings depending on the population added, the income level and other variables in a given time interval. With durable buildings, the model explained rents that rose with distance from the center and possible abandonment of buildings. Later, with Leon Moses at Northwestern, we published a two dimensional version of the monocentric city with public and private transportation competing and producing various land use patterns.
At Penn I discovered two schools of thought on modeling spatial interactions. Alan Wilson at Leeds, a geographer with a background in physics, had developed the macroscopic approach of entropy. Dan McFadden, an economist at Berkeley, was working on a type of econometrics particularly suited for urban studies because it incorporated random heterogeneity in a tractable way. Those who knew both approaches regarded them as inconsistent with one another, but I quickly realized the strong similarity and looked for a synthesis. This resulted in my paper on discrete choice, entropy and the equivalence between multinomial logit and the gravity models.
My true mission as a scholar was the development of computable models of city structure that could be used to make predictions and evaluate policy. Looking back, I have approached this goal systematically and holding myself to strict standards, but perhaps moving a bit too slowly. The standards are: (1) the empirical models should be an implementation of the theory itself preserving cause and effect relationships, not a distortion of the theory; (2) that as much of the available data as possible should be used; (3) that one should avoid over parametrization (a reason I never warmed to spatial econometrics and other descriptive approaches that place more emphasis on errors than on theory.)
My book published in 1982 introduced a multi-centric (as opposed to monocentric) version of location and rent theory based on discrete choice, with the aim of making urban economics empirically and computationally pliable. Using a version of this approach, I forecast how much residential rents near transit stations would change on the planned Southwest Corridor rapid transit line in Chicago. Later, after the Orange line going to Midway was built there, John McDonald and Dan McMillen measured actual rent changes in the Southwest Corridor and concluded that my predictions were accurate. This is the only case of validated prediction that I know in urban economics.
More recently, I have extended the approach to develop the “Regional Economy, Land Use and Transportation (RELU-TRAN)” model, a computable general equilibrium model of an urban economy published in 2007, and I have applied it to Chicago. Articles on the model’s application to Los Angeles and Paris are now in preparation, and there is a European project attempting to apply the model to Amsterdam, Barcelona, Goteborg and Istanbul.
Belief that the science should be applicable has always been the hallmark of my efforts, but I am often frustrated. Modeling cities should be the work of urban economists. They have the best theory of human behavior and of how to allocate resources optimally. Some excellent scholars from other disciplines, notably from geography, urban planning, and transportation, also model cities but with little or no economics. In the real world, such work competes with economics. Meanwhile, the “new urban economics” of the 1970s and 1980s produced much theory, but since the early 1990s there has been a furious flurry of interesting empirical work, adding relatively little to known theory. As such the anachronistic monocentric city is still often used to draw incorrect policy prescriptions. For example, an unregulated congested monocentric city sprawls excessively, but polycentric models such as those I have used show that more urban sprawl is often efficient. It does matter with what tools one looks for the truth.
(Published on RSAI Newsletter 2017 May)