Fixed link

Innovation, Organization and Economic Dynamics Selected Essays

Giovanni Dosi, Edward Elgar (2000)

posted on 15 March 2002

reviewed by Joe McCauley

As is pointed out by Paul Ormerod [1] in ?The Death of Economics?, neo-clasical economic theory self-destructed in 1968 when Roy Radner proved that infinite computational capacity would be required of rational expectations agents in General Equilibrium Theory [2]. In spite of that the truck keeps running without wheels, gaining in influence in law schools and in US court cases. Why this is so, how a theory without data (and, in eonometrics, data without theory) can survive and even prosper is a historico-sociological question, no easier to answer than are analogous questions about the persistence beyond the middle ages of Aristotelian theory, or the emergence of the Taliban in our own time. However, as the following quote (on microfoundations) suggests, Professor Dosi brings his criticism and ideas to us with exceptional wit and humor:

??I mean the construction of the dynamics of micro constituents of a system (be they molecules, ?, firms, cells,..)which yield the dynamics of some aggregate variables of the system?..the issue of microfoundations is in principle separate from that of ?microintentionality?, although obvioulsy in social sciences the relationship between the two is a tricky problem. However, contrary to the prevalent opinion amongst economists, I think it is healthy to keep the two matters separate. ? no physicist in a right mind would claim that molecules tend in probability to some energy state because they ?maximize their utility?, so some partial decoupling might also help investigators??

The theme of microfoundations of aggregate phenomena is taken up in chapters 12 ans 23. As Paul Ormerod [3] has taught us, neo-classical individual demand curves do not apply to aggregate phenomena other than maybe cornflakes (indeed, Osborne [4] has taught us that they do not even apply to individuals!). Furthermore, the neo-calassicists themselves proven that, given a textbook individual demand curve, the aggregate demand curves can be anything at all [3].

Professor Dosi?s autobiographical introduction tells us of both his intellectual trajectory (his indebtedness to Nelson and Winter) and provides the reader with a useful outline of the chapters in his (somewhat lenghty and, for a physicist used to a few equations, wordy?703 pages, medium print) book. And I must confess that reading the book (I have not read many chapters yet, so this is a pre-review) is made all the more pleasant with my memory of Giovanni and Company at late night sessions of the recent Econophysics Forum at Ovronnaz, with an adequate supply of bottles of wine on the table (and with earlier sessions in the hot springs before meals). I confess also to be impressed to learn that there are economists who know what ?computability? is about, and even write papers trying to apply the idea (how many physicists can tell you what ?computability? means?).

The reader in search of recent work on modern topics will find much food for though in the 23 chapters of reprints by Dosi and collaborators. Among the topics covered are evolutionary models, many chapters on path dependence, some on computability of models, and stylized facts (a phrase that I am still not comfortable with).

Regarding financial markets as tangential, not the central issue, the idea of economists is to try to understand how and why there is economic growth, and why there is a business cycle. This leads to a search for regularities (difficult to identify in the socio-economic field) or at least slowly-changing quantities that qualify as qiasi-invariants. After the 1968 crisis in economic theory two alternate ways emerged as possible microeconomic foundations: the dominant one got off the track in the search for invariance by introducing the notion of ?the representative agent? (approximate the system as one particle with no brain..) and variations thereof. On the other side of the fence were the evolutionary models, also trying to figure out what the analog of a particle or gene is (the firm? a subset of the firm? behavior invariances?), so to speak, in order to write down models based on the identified quasi-invariances. In the evolutionary models there arises the question how people learn, and therre are models of organization based on problem-solving (noting that typical risk behavior is the gambling tendency to throw good money after bad), which leads into psychology. How do markets emerge and work? What is the dynamics of organization? The text presents us with some stochastic models so, again, there is food for thought.

Nearly nothing is known about aggregates, making modelling there all the more difficult. The neo-clasical answers, Dosi correctly informs us, are too close to pre-Galilean reasoning (a theme that has been developed elsewhere, for good measure [5]). Also with foresight, Professor Dosi argues in several chapters that ?knowledge? is not equivalent to sheer information, which leads us into the topic of this era, the economics of information (here, he mentions the work of Arrow, Aoki, Ackerlof,, Stiglizz, Radner, Hurwicz and others). On increasing returns, path dependence and evolution we have

?The very nature of information, briefly?entails some form of increasing returns: the costs associated with its generation are basically upfront..can be generally reproduced and utilized at any scale with negligible marginal costs. This is so, for say, Pythagoras? theorem (whose ?production cost? was basically born by Pythagoras himself)?.knowledge is typically put to use under conditions of increasing returns..?

In Substantive and Proceedural Uncertainty (by Dosi and Egidi) a distinction is made between information incompleteness and knowledge incompleteness. Also interesting is that the authors recognize and emphasize the difference between the nature of probability in ordinary gambling and in economics. In the former cases the number and nature of bets placed cannot change the outcome: preferences, states of nature, actions and consequences are easily separable. In this case there is an information gap (we don?t know in advance wihich face of a die will show when tossed) but not a competence gap. In economics and finance, in contrast, the ?events? are not ?states of nature? but are an inseparable part of the decision process, are not independent of agenmts? actions. We read further that ?innovation implies nonstationarity..? and that ?rational procedures? may quickly run against the computational constraints of individual agents. E.g., Nash equilibria may ?exist? mathematically but may not be recursively realizeable (may not be algorthmically computable?for algorithmic computability in nonlinear dynamics, see [6]). On rational choice, they write that empirical evidence on individual behaviors is messy and subject to perturbations. Especially under uncertainty people may make mistakes anticipating expected probabilites and may not maximize. Even in 2-person games winning strategies exist mathematically but may require uncountably-much knowledge in the form of the infinite binary game-tree.

Finally, we learn that there are two extreme views in economics: the first is that there are no $100 bills on the sidewalk or, a bit more extremely, as a committee charged to decide whether to invite a rational expectations economist as professor once said, ?if it were a good idea he would already be here.? The other extreme, where Professor Dosi, our Giovanni falls, is that indeed there are $100 bills on the street even if we can?t axiomatize the process leading to their discovery!


1. P. Ormerod, the Death of Economics, Wiley (1997).

2. Radner, Roy (1986) 'Can Bounded Rationality Resove the Prisoner's Dilemma?', in Hildenbrand, W. and Mas-Collell, A. (eds.) Contributions to Mathematical Economics in Honor of Ge/rard Debreu (Amsterdam: North-Holland), pp. 387-399. [Uses finite automata.];

3. P. Ormerod, private comment (2002).

4. M.F.M. Osborne, The Stock Market and Finance from a Physicist?s Viewpoint, Crossgar (1996).

5. J.L. McCauley, Physica A237, 387 (1997).

6. J.L. McCauley, Chaos, Dynamics and Fractals, Cambridge (1992).