Ormerod’s Positive Linking
Paul Ormerod Positive Linking: How Networks are Revolutionising Your World Faber and Faber, 2012.
In The Death of Economics … I argued that conventional economics views the economy and society as machines, whose behaviour … is ultimately predictable and controllable. …
In Butterfly Economics … I analysed a wide and seemingly disparate range of economic and social questions, seeing them as analogous to living creatures whose behaviour can be understood only by looking at the complex interactions of their individual parts. …
Why Most Things Fail [demonstrated] close parallels between … the extinction of biological species … and the extinction of companies. …
[These last two books] were fundamentally based on the concept of networks. …
The assumption that people make choices in isolation, that they do not adopt different tastes or opinions simply because other people have them, is no longer sustainable. [Our world] is completely different from the world of conventional economics … . It is the implications … which I explore in this book.
Most public policy on social and economic matters is based on the premise that people, or indeed companies, behave as individuals when they make their decisions.
When the power of the network takes over, people are no longer acting autonomously, but as part of a social group, and their behaviour and decisions are driven by the process of copying, of imitation.
[The] economic definition of rationality which has become irrational.
Ormerod has a good example, extensively discussed, of police action triggering a riot. He notes that in economics the term ‘incentives’ is used in a particular way, such that it seems debateable as to whether people should, or should be expected to, respond to incentives. In contrast …
Network theory allows the social dimension of human activity to be taken into account … .
Thus, Ormerod argues that people do not have fixed preferences, but change them in response to the actions of others.
Up to a point, Lord Copper
Successful policy-making in the twenty-first century requires an understanding of networks and incentives , … . [But first,] it is worth considering briefly whether the use of abstraction and of maths can be justified in analysing social and economic problems. … I have no problem at all with abstraction … . [Meaning the selection of factors.] It is rather harder to justify the use of mathematics … , especially the specific sort favoured by economists.
[Maths] is a tool which can assist in logical thinking. It’s like another language. It can help us find our way around and serve as a medium of communication amongst people discussing the same subject. [On] the one hand, there are good reasons for the use of maths in economics, and on the other there are bad ones. So far, mainly the bad ones have prevailed.
[A] particular kind of maths … is inextricably linked with the concept of Rational Economic Man [i.e. differential calculus]. … This is still the basis for a lot of the economics which is taught today. Yet, paradoxically, it has been precisely the use of maths itself within economics itself which has exposed fundamental problems at the very heart of the Rational Economic Man theory of behaviour.
Ormerod cites some maths to show that, counter-intuitively, one can have increasing overall sales with price, even if every economic agent is ‘rational’ in the sense that their demand decreases with price (e.g. if increases in price are passed on to labour or share-holders, who are then able to buy more.) (This is related to the observation that austerity, tax increases, attracting the super-rich, etc, can all have counter-intuitive impact, depending on how money actually spreads through the economy. Mathematics shows that here is no fixed rule, and shows what one needs to pay attention to.)
[In] most real-world social and economic situations, we need to understand the potentially subtle and powerful interplay between the effects of incentives and the effects of networks.
The standard response by economists is to invoke the ‘as if’ argument.
[The] mainstream ‘rational’ agent argument [has it that] over time people learn to avoid decisions which are not in their own interests. [But] in many situations, the best choice can never be identified, even after the event, no matter how smart we may be.
Ormerod discusses the falling trend in crime. He regards the study results as inconsistent with optimising behaviour, but consistent with his notion of incentives and linking.
The Shoulders of Giants: Keynes and Simon
Ormerod discusses Herb Simon’s work. He quotes Simon:
“[The] task is to replace the [assumption of] the global rationality of man with a kind of [bounded] rational behaviour which is compatible with access to information and computational capacities actually possessed … .”
Ormerod discusses behavioural economics:
[In] many ways it sidesteps the most fundamental challenges which Simon posed … . [In] many real-life situations, the optimal situation can never be known. … Choosing a partner, selecting a pension scheme, deciding whether to take another job, casting a vote in an important election, … .
He discusses the use of rules of thumb:
Agents use them as long as they continue to deliver … satisfactory outcomes. [Often] I simply do not know which of the alternatives might have been the best … .
He quotes psychologists:
“As humans, we generally have the feeling that we decide what we want and what we do. … We can envision ourselves in … alternative futures … . We only have to decide to do so, and we can go and see a movie … or hang out with our friends. It is up to us. Our behaviours seem to originate in our conscious decisions to pursue desired goals, or outcomes.”
Ormerod quotes Keynes’ anticipation of Simon:
“We have devised for the purpose a variety of techniques [rules of thumb] , of which the most important are … :
- We assume that the present is a much more serviceable guide to the future than a candid examination of past experience would show it to have been hitherto. In other words, we largely ignore the prospect of future changes about the actual character of which we know nothing.
- We assume that the existing state of opinion as expressed in prices and the character of existing output is based on a correct summation of future prospects, so that we can accept it as such unless and until something new and relevant comes into the picture.
- [We] endeavour to conform to the behaviour of the majority or average. …”
In short, Keynes argues that copying the behaviour of others is … the ‘conventional judgment’.
Did Economists Go Mad?
Ormerod discusses various crises since 1997.
The problem appeared solved and the world moved on. …
‘The state of macro is good.’ The state of macro is good! In August 2008!
[No] one really understood the level of risk that was being carried at any one time. … In Keynes’ phrase, the bankers were ‘a society of individuals each of whom is endeavouring to copy the others’. if they copied the opinion, the belief, that everything was fine, it would continue to be so. … But once they stopped believing, we had a credit crunch.
Pessimism spread like wild-fire.
Ideas such as [those] at the heart of modern macroeconomics have provided the intellectual justification of the economic policies of the last ten to fifteen years. And it these ideas which the recent crisis has shown to be false. The dominant paradigm … has been that of rational agents … using rational expectations. … The requirement is that on average, over a long period of time, such expectations are correct. … An assumption of the theory is that … the agent is in possession of the correct model of the economy. .. DSGE models … ruled out the very feature which distinguishes almost all financial crises: … loss of confidence … .
Policies which generate confidence are … not so much … the economically ‘rational’ calculation of their positive consequences, but the creation of a positive mind set. … Positive linking.
Lady Luck: The Goddess of Fortune
Careful, practical study is needed on a case-by-case basis to determine whether a free market is likely to lead to stable or unstable behaviour.
Ormerod discusses work by Duncan Watts, showing that both robust and fragile networks can emerge from the same initial conditions.
[Networks] are becoming more and more important in the social and economic world.
The World is not Normal
The normal distribution underpins almost all the technical statistical analysis carried out by economists. [But] Herb Simon … noticed that across a very wide range of examples, data was distributed in a completely different way. (E.g.,] distribution of incomes … .
Ormerod discusses Albert-Laszlo Barabasi’s work on ‘preferential attachment’, scale-free networks, small worlds, etc.
[Where] networks are important, generalised statements about policy effectiveness across different policy domains lack validity because of the different natures of the different types of network which predominate in the social and economic worlds. …
Hayek had earlier noted that ‘it is perhaps the most characteristic feature of the intellectual that he judges new ideas not by their specific merits but by the readiness which they fit into his general conception … .’ …
Copying is the Best Policy
All theories are approximations to reality … . [In the conventional theory, Rational Man] operates independently, not being influenced directly by other agents, with fixed tastes and preferences.
[‘Copying’] has become the rational way to behave … . [This is] the mode of behaviour in which your choice is being influenced , altered, directly by the behaviour of others.
Ormerod discusses attempts to improve rice yields in Bali. He also reports a ‘social learning’ tournament. Copying was a good strategy.
However … using copying as the main way of making choices can lead to outcomes that are so radically different from those predicted by rational economic behaviour that it really does not make sense to describe the former as ‘rational’ in the economic sense of the word at all.
There is an extensive discussion of music downloading, which conflicts with supposed rationality. Also of Brian Arthur’s example of drawing balls from an urn, always adding back an extra ball of the same colour as that drawn, as a demonstration of emergence.
All Things Must Come to an End
Herb Simon’s paper on skewness [described] the basic mechanism which underlies the emergence of unequal outcomes we observe in the real world of human social and economic systems in a wide range of disparate settings. …
Rational behaviour … is essentially based on … ‘copying’ .. .
Ormerod discusses Mike Batty’s work on city sizes. This is not consistent with preferential attachment, but has a ‘micro-dynamics’ that gives ‘churn’ in city rankings. He describes a variant of Arthur’s urn model, in which with some small probability a ball with the wrong colour is added in. This is more realistic. He goes on to consider the range of examples considered when thinking about what to copy. This makes things much more realistic still.
What Can be Done
Keynes argued that the very concept of rationality needed to be redefined in such a world.
[Public] policy interventions … have been based on the assumption that agents respond solely to incentives.
The principle cause of failure of what we might describe as the social democratic model to achieve its objectives is not the size of the state but the intellectual framework in which it operates. [The] world is seen as a machine … . It is a world in which everything can be quantified and targets can be not only set but also achieved, thanks to the cleverness of experts.
Rejecting the model [of the previous chapter] because it does not give unequivocal policy recommendations is wholly invalid.
[Agents] consciously choose to copy as the basic principle of their behaviour in many situations. It is entirely rational for them to do so.
[When] network effects are present, the most effective policies are unlikely to be generic, across-the-board changes to incentives. … Altering the structure of the network might itself be also become a policy target, and one which could have powerful effects. … [A] reliable indicator of whether network effects are present [is that] we observe unequal outcomes.
[What] really matters is how persuadable are the agents who are directly connected to the agents who initially alter their behaviour.
[We] need systems which exhibit resilience and robustness, which can respond well to unpredictable future events, which can recover through the strength of positive linking.
Communities ought to take responsibility for outcomes … . [There] is no silver bullet.
Let us take as an example the management of economic crises. The rational paradigm has spectacularly failed to explain and predict not just how the present crisis emerged, but how it spread and burgeoned …. .
Inequality is a major concern … . [Also] the ‘tragedy of the commons’.
We do not yet have the answers to such shortages, or many other major problems, but an approach based on [positive linking] is more likely to deliver success ….
Ormerod supports the view that neoclassical economics was never realistic at the micro level and that while it may have once been realistic at the macro level the rise of networks makes the theory untenable, and the financial crisis from 2007 provides adequate evidence of the inadequacy of it for the current situation, particularly for government policy. An uncritical belief in any false theory will typically lead to error, and Ormerod identifies some specific issues and recommends some areas for improvement drawing on network theory. But it is not clear to me that he always shows how these impact at the macro level through policy. But then the neoclassical theory is confusing, and Ormerod does not establish a clear alternative framework. It is conjectured that a better theory would assist in tackling inequality and various tragedies of the commons, but this is work to be done.
One problem I have is that neoclassical economics in effect assumes that the economy is relatively closed and stable. This would seem to imply that policy can have little impact. Without the benefit of neoclassical thinking one might suppose that economies were potentially unstable and that it was a policy matter to avoid crises. But the neo-classicists assure us that no-one is needed to ‘mind the shop’. The other potential role for policy would be to move to another equilibrium (e.g., with less inequality). But neoclassicism tells us that this is not possible, and if it were then neoclassical theory would not apply. There might be roles for government be in smoothing out variability (via changes in interest rates) and enabling mega-projects (such as infrastructure), but in effect the neoclassical theory determines the policy, which the government then enacts.
One way to look at this is in terms of short versus long-term thinking. As Ormerod says, most people act according to their habits most of the time, as if they are automata with no free will, and they typically only change their habits when there is a problem. The neoclassical view is that rational man may make some mistakes, but learns from them and converges on some effective habits that are based on concrete knowledge. It seems to me that the neoclassical view is correct in so far as the functioning of a closed economy would be based on ‘knowledge’, including the knowledge of potential crises, and thus able to avoid them – if that was what was wanted. Hence I suppose that not enough people did have the necessary knowledge, and that what we want to avoid future crises is more knowledge. So what I imagine is that people learn habits which are effective in the short-run, based on knowledge which has not been falsified. Thus neoclassical economics may be broadly right ‘in the short run’, but no longer.
We lack a theory of the longer-term. Ormerod does give us some insights, but first I want to mention game theory. The assumptions of neoclassical theory assume that there is a fixed game, with fixed coalitions among players. Changes in the rules of the game or of coalitions will tend to lead to changes in networks, and change in networks can lead to changes in the games. Either game theory or network theories can be used to point out the limitations of the neoclassical view. The network view has the advantage that it is less different from the neoclassical view of individual agents than the game theory view. But it seems to me that once one accepts that there are policy aspects then a multi-level game view is more appropriate, or at last should be considered as well. For example, the neoclassical view supposes that agents make decisions independently, and this can still be true in networks. But game theory emphasises the role of coalitions, where people can do better by coordinating activity. This is particularly important where people don’t just value money.
From a game theory perspective, once a stable regime has set in it is understandable that agents should settle into familiar games, until the stability is challenged, such as by changes in the structure of economic power. Ormerod notes that ‘mathematical’ models of ‘rational man’ use differential calculus, and seems to suggest that such mathematics is somehow ‘bad’ or at least inappropriate. Yet Turing used such mathematics to show that if one adds in some randomness to a system of differential equations that are ‘relative’ (i.e. without any pre-determined solutions) then not only does one replicate the qualitative behaviour that Ormerod describes in networks but one also has ‘critical instabilities’. The idea here is that one can have a continuous evolution in which the randomness has negligible effects most of the time, but where from time to time a small input can have a substantial long-term impact. Typically, the system can be moved from one pseudo-equilibrium to another. Sometimes the system can move rapidly from one pseudo-equilibrium to another, never settling down and appearing not to be in any equilibrium. It seems to me that this model supports the need for policy, to prepare for possible instabilities, to look out for up-coming instabilities, to determine preferences across possible futures, and to support action. Within this, rational man would optimize as before, but would give some utility to looking ahead, and more utility when the situation appeared potentially unstable. This seems to me more realistic than the neoclassical theory but less of a change than using network or game theories. It may be what Keynes was recommending, above.
Ormerod recommends ‘copying’, by which he seems to mean looking at what others are doing and adjusting one’s own beliefs and activity accordingly. If every generation has a Keynes or Simon then this may be adequate. But how would we know who to copy? Or when to change? If we are a small player, then copying would seem to be a good investment strategy: we get in and bail out early, maximizing returns. This is what many were trying to doing up to 2008. But copying seems inadequate for policy-makers. The advice is to ‘alter the structure of the network’, but how? The game theory perspective seems more useful at this level. One might seek to ‘change the game’, considering a range of options, not just changing the network.
Ormerod is critical of policies’ focus on ‘incentive’, and one could characterize game theory as about incentives. But changing the game is changing incentives more radically than either simply adjusting pay-offs, e.g. through tax and interest rates, or changing networks.