The Evolving Web of Future Wealth
The web of connections among goods and services in an economy may be the long-missing key to understanding how novel innovations and new wealth arise
By Stuart Kauffman, Stefan Thurner and Rudolf Hanel
In Brief Traditional growth theory ignores the importance of the "economic web" of connections among various goods and services within an economy. In particular, economists have failed to recognize how much the structure of this economic web does to drive innovation and opportunities for future wealth. Successful innovation depends in part on bringing together goods and services in useful, unanticipated combinations. Because all the possible useful combinations can never be defined in advance, it's impossible to predict the future trajectory of innovation. Still, by analyzing the structure and diversity of an economic web, it's possible to determine whether one is "collectively autocatalytic" and will spontaneously expand to encompass additional economic niches. "Gales of creative destruction" blow through economic webs, eliminating some old goods and services as novel ones emerge. Details of a web's structure may determine how dynamically it balances between growth and collapse. The study of economic web behavior may offer insights into why some real-world economies have such difficulty prospering.
Editor’s Note: Stuart Kauffman has a well-earned reputation as a scientific provocateur, albeit one with the weight of data and wisdom on his side. Kauffman, a complexity researcher and biologist of the University of Calgary and the Santa Fe Institute, has argued, for example, that self-organization—the propensity for systems to become more complex without outside guidance—was just as important as natural selection in shaping evolution. (Intelligent design advocates, take note.)
In his new book Reinventing the Sacred: A New View of Science, Reason, and Religion (Basic Books, New York; May 2008), Kauffman develops a larger argument: Understanding what’s happening in complex systems could help modern science break free of what some consider its too-reductionistic underpinnings. One controversial idea that Kauffman develops in his book is that by failing to take this approach to economics, traditional economists are unable to explain something that seems obvious but isn’t: How does innovation drive growth?
In this article, a rough draft of which appears below, Kauffman and his colleagues Stefan Thurner and Rudolph Hanel detail some of their thinking on the subject, which is sure to raise the hackles of some members of the economic community just as their ideas on biology have ruffled some scientific feathers.
- What are your reactions to their arguments here?
Do you think economists have not formally incorporated innovation and its relation to the growth of wealth into their theories? If so, why?
What are your thoughts on the proposed “grammar model” approach to describing how goods and services evolve as economies grow?
Tell us why you agree or disagree with the idea that conventional economic theories fall short and that a new theory explaining the relationship between innovation and wealth will be economically useful?
Give us your answers to these and other questions raised by this provocative piece in our comments section below. Your feedback will be incorporated into a version of this article that will appear in a future print issue of Scientific American.
Economies have existed for as long as humans have relied on artifacts and trade. Fifty thousand years ago, the estimated diversity in the global economy might have been between 100 and 1,000 types of goods and services. Today, economist Eric Bienhocker of McKinsey and Company estimates the diversity of goods and services available in New York City alone at about 10 billion. Perhaps the most stunning feature of the economy over time is the explosion of goods and services. Yet contemporary economics has no adequate theory to understand this explosion or its importance for economic growth and the evolution of future wealth.
Economic growth theory is highly sophisticated about the roles of capital, labor, human capital, knowledge, interest rates, saving rates and investment in existing economic opportunities, or investment of savings in research to find novel goods and services. Yet the major conceptual frameworks that undergird contemporary economics (competitive general equilibrium, rational expectations and game theory) share a crucial failing. They assume that all the goods and services (as well as the relations between them) and all the strategies for engaging with them in a local or global economy can be “pre-stated”—that is, known in advance. In reality, novel goods and services may constantly enter markets, thereby requiring economic actors to develop ever more novel strategies: all the relevant variables cannot be pre-stated.
Thus standard growth theory misses an essential feature of this “economic web” of goods and services. Even more important, as we shall explain, it ignores the role that the structure of the economic web itself plays in driving the creation of novelty and the evolution of future wealth.
Here is a simple example. Had a firm invested vast sums of money in 1910 to invent the television remote control—many years before the invention of the television, multiple channels, substantial programming and a substantial viewing audience—the TV remote would not have fit into any niche in the existing economic web. The invention would have been useless and unprofitable; it would have propelled neither investment nor the creation of wealth.
That case in point tells us that the structure and details of the economic web are themselves preconditions of the invention and investment that will effectively drive wealth creation. Furthermore, the novel good that fits into an economic niche thereby extends the economic web in new ways—creating new niches in what we call the “adjacent possible” of the economy and thereby inviting further invention, investment and wealth creation.
We tell a story that we hope is true (but is in any case illustrative). Some engineers were trying to invent the tractor. They knew they would need a massive engine block. So they mounted a massive engine block on a chassis… which promptly broke. They tried a succession of larger chassis, all of which also broke. At last one of the engineers said, “You know, the engine block is so big and rigid that we can make use of that rigidity and hang the rest of the tractor off of it. We can use the engine block itself as the chassis.” And, in truth, that is how tractors are made (how formula racing cars used to be made).
The rigidity of the engine block that allowed it to be used as a chassis was the economic equivalent of what Charles Darwin called a “preadaptation” in biology. That is, rigidity was an unused causal feature of engine blocks, suddenly put to a novel functional use. Engine blocks are rigid for reasons that have nothing to do with their potential to become chassis, but because that rigidity existed, it preadapted them for that use. We believe such preadaptations are common in the evolution of the economy.
As another example, consider the modern linkage of computers, first invented to solve projectile trajectories in World War II, into a globe-spanning Internet. No one foresaw the Internet in 1948; the computer was nonetheless preadapted to that novel communications functionality.
The ways goods and services come to be used together in the real economy may not be describable in advance either. Cell phones were first introduced purely for voice communications; then text messaging emerged as a simple additional innovation. In Japan, where text messaging is free, people now write all kinds of documents that way, adding to the gross domestic product. Some best-selling novels there were written on their phones by schoolgirls.
Because it is impossible to identify all the preadaptations and potential economic uses for goods and services, it is impossible to finitely prestate all the possibilities for them. This conclusion has profound significance: it means that predicting future innovations is fundamentally incalculable, even on the basis of probability because no probability distribution can be assessed without knowing the range of possible outcomes. (And beyond economics, this principle may have equally radical consequences for much of the rest of science [see sidebar].)
Decision theory—the tool of management that suggests making optimal choices by summing discounted future values over the probability distribution of all possible outcomes—is of limited usefulness, as are businesses’ five-year plans. We cannot deduce our lives; we must live them forward, as the philosopher Soren Kierkegaard said, even when we face unknowable uncertainty. Business, like life in general, is an art wherein we must use reason, intuition, emotion, metaphors, models, case studies and more to guide ourselves. Business is not a calculus. Thus, economics can only partially be a calculus, and a much broader conceptual framework is needed.
The Economic Web
A screw and a screwdriver are complements, used together to create value by, say, fastening two boards. A screw and a nail are largely substitutes: loosely speaking we can use one where we use the other. Now imagine all 10 billion or more distinct goods in the contemporary global economy as points in a large three-dimensional space. Join complements by green lines and substitutes by red lines.
This network is the economic web. We do not know its structure, but it exists. It evolves over time, although we know little about how. Do statistical laws govern its evolution? Are firms located near the center of the web (say, automobiles or computers) in a strategic situation different from those on the periphery, such as hula hoops?
The web of complements to a good forms a mutually self-reinforcing and cross-reinforcing subnetwork that enhances its own economic growth. For example, with the car came its complements, among them gasoline, paved roads, motels, fast food restaurants and suburbia. In turn suburbia gave rise to an enormous number of consumers of automobiles, gasoline, paved roads and so on. We might call such mutual cross-enhancement “collectively autocatalytic,” in that each component helps create the economic environment and market for the others and all mutually benefit. In economic terms, we might call them collective webs of mutually positive “externalities” between complementary technologies. Such collectively autocatalytic webs of complements can drive enormous wealth production, provide a very large investment incentive and massively promote the evolution of future wealth possibilities.
In contrast, consider the hula hoop, which appears to have few complements. It may have made money for its producers, but it sparked no lightning in the form of complementary technologies or products that collectively drove an explosion of wealth. The hula hoop could come and go with little effect on the economy.
The preceding observations show why we must come to understand the structure, evolution and roles of the economic web. Of course, it is people who invent novel goods and services, but the structure of the web itself singles out where invention and investment are likely to yield a profit and drive growth.
Two further features of the web make us suspect that the diversity of the economic web drives its own growth autocatalytically. First, consider the Wright brothers’ airplane: fundamentally, it is a recombination between an airfoil, a light gas engine, bicycle wheels and a propeller. The more goods and services that exist in the economy, the more recombinations among them are possible. Place an umbrella down the smoke stack of the old Queen Mary and you achieve a mess. Place it behind a landing Cessna and you have invented an airbrake.
Second, new goods and services typically enter the economy as complements or substitutes for existing goods and services. Call the set of goods and services that are complements or substitutes to a given good or service its economic niche. As the web grows, does it create new niches faster than it creates new goods and services? The general answer is not known, but the very large number of complements to the automobile and computer noted above, with their mutually cross enhancing externalities, suggest that the average number of new adjacent complements and services created per new good or service is greater than 1.0.
If so, then the growth of economic niches is indeed autocatalytic. The more goods and services that exist, the higher the diversity of the economic web and the faster the creation of new economic niches. Thus the very diversity of the economic web is almost certainly a major factor in creating the conditions for its own further expansion.
We do not yet know whether that is so, nor whether the average number of novel niches created per new good has changed since 50,000 years ago. Economic historians can discover the truth. But in the meantime, we note that these issues are not yet part of economic theory, and may be major, largely overlooked factors. If so, they may have practical implications and deserve detailed examination.
An Algorithmic Model
There are profound reasons that the structure and growth of the economic web is not part of current economic theory: modern economic theory is deeply mathematical. What mathematical framework would allow us to say that the screw works complementarily with the screwdriver to create value? What algorithmic model can describe unforeseeable Darwinian preadaptations in the economy? There may be none.
The hope of finding a mathematics that could describe and predict how novel goods and services unfold as the economy evolves into its adjacent possible thus seems precluded, at least at present. But even if the growth of the economy is not algorithmic, an algorithmic approach may still be of use in finding statistical features of model economies for comparison to the real one. Crucial here is the enlargement of the current framework: a concept is needed to mathematically tame the “adjacent possible.”
One such approach is a “grammar model” that represents goods and services with binary symbol strings, such as (000). Within our model, the number and diversity of the strings can stand for renewable resources, appearing each year. Symbol strings can act on one another to create new symbol strings. For example, a symbol string with a (000) in it can rewrite the (000) in a second string into a (1010). A “grammar table” lists all the pair rules for these transformations. This arrangement can simulate a simple economic production function.
Intuitively, one sees that if the starting (and renewable) number of strings is small, that their diversity is low and that the grammar table has few pair rules, symbol strings will probably not be able to act on one another and few novel symbol strings will be created. We call such behavior subcritical. A subcritical economy cannot generate a growing diversity of goods and services. On the other hand, studies show that as the number of pair rules, resource strings or both increases, the system can abruptly transit into a supercritical domain where a large—perhaps unending—diversity of symbols strings may be generated. We call this explosion of goods and services supracritical.
Networks of Productive Pairs
We have recently idealized the above model and have confirmed analytically and numerically the existence of the subcritical and supercritical phase transition. In this idealization we map the problem of interacting strings into the setting of the autocatalytic networks that describe how one good can act on another to produce some third good. We will call such pairs productive.
We can take a wheel and a rope, for instance, and combine them to form a tackle, or we can use the rope to secure a boat at the pier. So ropes can be used in various productive pairs, as in (rope, wheel) -> tackle. Obviously most pairs we form randomly will not be productive. There is no rule that would allow us to do something useful with a supernova and a fish (except possibly in some psychedelic science fiction novel).
Next, we can ask whether the underlying production network instantiates some good, beginning solely with a fraction of the available possible goods. In a way we mathematically benefit from our profound ignorance of the real economic web’s detailed structure because it forces us to model the catalytic network as basically random.
The absence of any particular plan underlying some catalytic economic network allows us to see something fascinating. If a random catalytic economic network contains sufficiently many productive pairs, then below some critical number of initial goods there are insufficiently many productive pairs to sustain the invention of new goods. Above this critical number of initial goods, however, practically all possible goods arise within comparatively few generations of production.
Some real-world economies appear to be subcritical. Joseph E. Stiglitz of Columbia University describes one African nation whose major economy consists only of diamond and cattle exports. One of us (Kauffman) lives in Alberta, Canada, which exports shale oil, animal and forest products, and has an information technology industry correlated with the oil industry. These two economies appear to be subcritical: they do not seem to be creating an ever growing and changing diversity of goods and services complement and substitute for one another. By contrast, the U.S. economy, the European economy, the global economy and perhaps other national and regional economies appear to be supracritical, creating an ever changing spectrum of novel goods and services.
Growth opportunities for subcritical and supracritical economies may be very different. Supracritical economies endlessly afford new economic niches that invite invention, local and foreign investment and the potential for great wealth creation. As old technologies die, capital can migrate to new sectors of the supracritical economy to drive further wealth creation. Subcritical economies appear not to have these properties. For instance, in Alberta, oil wealth currently yields a $7 billion (Canadian) surplus annually. But it seems reasonable that in several decades, alternative energy resources will be more than competitive and preferred by ecologically conscious consumers. Demand for Alberta oil may then decrease. As it does, where locally will capital migrating out of the dying oil industry go? In a subcritical economy, no major new investment opportunities may exist.
The speculative possibility that subcritical and supracritical economies have very different growth properties seems worthy of study. The persistent diversity of wealth in nations across the globe is a puzzle to many economists. If a distinction between subcritical and supracritical growth patterns has merit, what are the implications for attracting local and foreign investment into poverty-stricken subcritical economies around the world? If 45 percent of the world’s population lives on a few hundred dollars a year, it seems worth exploring the implications of economic webs for economic growth. We may discover alternative ways of abetting economic growth.
Gales of Creative Destruction
The economist Joseph Schumpeter is famous for having introduced the idea of “gales of creative destruction.” When the car replaced the horse, many old goods and services (such as horse-drawn buggies, saddleries and smithies) went virtually extinct, while whole new industries and communities grew around oil, gas, paved road, motels and suburbia. With our algorithmic model we are able to quantify Schumpeter’s gales:
Let us return to the network of productive pairs we were considering earlier. Essentially all goods that are technologically possible are producing each other in self-sustaining patterns. There is no intrinsic reason why this pattern should ever break down. However, our model does not yet contain utility, which we may model as an external selection process. If then, as in human societies, some products go out of fashion because preferences shift (say, from horses to cars), then occasionally some goods will be cancelled in the fully self-sustaining network. We call such exogenously triggered extinctions “primary defects.” With each such defect, about twice that many productive pairs with which this good was involved will also vanish on average. As long as a good gets produced by at least one productive pair it remains in the game. When a good loses all its productive pairs it too will vanish. This disappearance we call a “secondary defect.”
It is intuitively clear that when the first primary defects are introduced into the system, the effect on secondary defects will be marginal. Most of the products born of defective productive pairs can still be created by alternative productive pairs. But at some point many goods can only be produced by a single productive pair, and all of a sudden a few more defects have devastating consequences. The whole system ceases to be self-sustaining and breaks down in a cascade of secondary defects–a gale of destruction.
Our idealized grammar table model yields Schumpeterian gales of creative destruction whose size follows a power-law distribution. Surrogate data, in the form of firm failures [over time?] in many countries, also follow a power law that is close to our predictions.
The basic conclusion is that catalytic economic systems have three dynamical regimes. In the sub- and supercritical regimes, the system is dominated by destruction or creation; although the system fluctuates, it essentially remains either in one or the other. However, when the system is prepared critically, we observe a different behavior wherein Schumpeterian gales of creation and destruction occasionally throw the system from one phase into the other.
Conventional economic theories about growth and the evolution of future wealth may be inadequate. We need a theory and historical examination of the growth of the actual economic web and of whether, in a supracritical economy, a sufficiently high diversity of the web autocatalytically drives its own growth. Furthermore, we need to understand the mutually and collectively cross-enhancing power of complementary technologies, regulatory structure and attraction of consumers in the creation of wealth. That understanding may lead to new views of how to foster economic growth worldwide, particularly in the poverty-stricken parts of the globe.
One potential implication of our algorithmic model may be that isolated or small economic regions may be foredoomed to remain subcritical, with more limited economic growth opportunities than supracritical economies. For those regions, union into larger economic domains, thereby increasing the diversity of goods and services (and of the economic web itself) as well as the aggregate market, may help yield supracritical growth. (Europe seems to have followed this path.) Understanding more about the statistics of Schumpeterian gales of creative destruction may help guide future research, invention and investment more wisely towards emerging large, mutually enhancing sectors of the economy. We hope that these concepts are plausible enough to warrant detailed investigation into their policy implications.