Societies as Complex Systems

May 20th, 2006

The theory of complex systems has captured the imagination of many people who think about the past and future of societies. The notion that societies may be thought of as Complex Adaptive Systems (CAS) is certainly compelling. CAS are made up of many interacting elements, each following its own local behavioral rules, which includes an ability to learn. A frequently-cited feature of CAS that is relevant to social dynamics is that CAS can shift abruptly from one regime to another. However, writing as a policy analyst who studies socioeconomic scenarios, this feature does not provide much guidance when thinking about societies. Instead, I would propose that the most important lessons to learn from CAS are that structure is not fundamental, and is inherently fuzzy.

Some very brief background:

In CAS, structure arises from local dynamics, so the structure is not fundamental – the local dynamical rules are. When the system settles down into a steady pattern, it moves rapidly between a large number of qualitatively similar states. Having a large number of states helps to stabilize the system and permits it to evolve, so fluctuations are not only unavoidable, they are crucial for the viability of the system. Hence structure is not fundamental in CAS, and is inherently fuzzy.

Since structure is not fundamental, CAS have one structure rather than another partly for historical reasons and partly because of additional factors that influence the system. Any structure consistent with the underlying dynamics and the prevailing additional factors is possible. As the influencing factors change, CAS can exhibit abrupt shifts from one qualitative regime to another. (In the case of societies or communities, an example of this would be a shift from extensive to intensive cultivation as population density increased.)

There are several reasons why the “abrupt change” feature of CAS is less compelling than the more pedestrian idea that structure is not fundamental, and fuzzy. First, in many cases, the assumption that the structure is fixed and unchanging is an excellent approximation. While it is important to bear in mind that the structure can change (because it is not fundamental), it is nevertheless analytically useful. Second, the meaning of “abrupt” depends on the relevant time-scales. In the case of societies, an “abrupt” change may take more than one generation to play itself out. Third, the particular parameter value that might trigger a change is unknown for social systems. Without knowing where the threshold lies and how long the change can take to occur, the “abrupt change” picture is not analytically fruitful. Fourth, much of what people experience in their lives as wrenching change is actually within the bounds of normal fluctuation. It is common enough to suppose that some event has “changed everything,” and in some sense that may be true, but not in the sense that society has shifted from one qualitatively distinct state to another. Many social structures are remarkably stable. Fifth, and finally, it is very difficult to say what structures are compatible with the underlying dynamics of the interacting parts of society, because these are very imperfectly understood. For all these reasons, the idea that societies and communities, as CAS, may exhibit abrupt changes in structure is not very informative for practical analytical work.

Starting instead with the idea that structure is not fundamental and is inherently fuzzy can, on the other hand, inform practical work. It leads to three concrete recommendations for the analyst:

  1. Do not take existing social structures for granted.
  2. Look for historical examples of change in social structure as part of your analysis.
  3. Accept variability as both inescapable and necessary for the viability of societies and communities and incorporate it into your analysis.

These observations are less dramatic, but perhaps more useful, than the picture of CAS leaping from one state to another.

Contested and Uncontested Stabilizing Factors

May 20th, 2006

A criticial question when thinking about societies is the question of order – why do societies maintain their structure over long times? Factors contributing to order can be placed within one or another of two broad categories – whether they operate when scarce resources are to be allocated (that is, in contested situations) or in day-to-day (uncontested) exchanges. Examples of factors that operate when scarce resources are to be allocated include:

  • Market forces
  • Categorical inequality (such as gender discrimination)
  • Political negotiation
  • Force

In contrast, factors that operate in day-to-day exchanges include human social norms and the “moral resources” (a concept described by Jonathon Glover in his book Humanity: A Moral History of the Twentieth Century).

A further factor, which falls into a gray area between these two categories, is the moral exhortation of classical civic republicanism. The citizen is expected to regulate all of his or her exchanges – contested and uncontested – in accordance with the precepts of virtue. However, as Thomas Hobbes argued, in practice the claims of virtue are weak when scarce resources are to be allocated (aside from the problem of negotiating between different visions of the good life), while Adam Smith argued that within an appropriate institutional framework, individuals’ vices could lead to a seemingly virtuous public outcome.

Both contested and uncontested factors are important. However, as Hobbes and Smith were arguing, when scarce resources must be allocated, basic human responses are often put to the side. When considering societal dynamics, it is important to look at the stabilizing forces in contested situations, since it is in these situations that the temptation is greatest to act in an unjust way. Adam Smith’s remarkable insight was that in many cases, selfish behavior could lead to socially positive outcomes.

In this connection, the ideas of Charles Tilly, as expressed in his book Durable Inequality are of interest. Tilly’s explanation of the persistence of durable categorical inequality – for example, inequality correlated with race or gender – is a challenge to market optimists. His proposed mechanisms operate independently of, and parallel to, market allocation mechanisms. Even in the presence of a well-functioning labor market, categorical inequality can prevent people from achieving the status they might otherwise have achieved if they were not members of an unfavored group.

In Durable Inequality, Tilly argues that within institutions (such as businesses) an effective way to allocate scarce resources (such as managerial positions) is to assign them using categories that are recognized in the broader society. For example, if broader society views men and women as different (and women as generally inferior), then it is relatively costless for the institution to borrow that categorization and largely exclude women from upper management positions. He also argues that the unequal structure is more stable if there is some leakage – that is, if there are some women in upper management positions, but not many, then it is easier to maintain the unequal distribution than if there were none. The main stabilization mechanism is opportunity hoarding, whereby the lucky few hold on to their gains at the expense of those still excluded.

The first 5 of the 8 elements of Tilly’s own summary of his argument for the roots of durable categorical inequality make these points:

  1. Paired and unequal categories…recur in a wide variety of situations, with the usual effect being the unequal exclusion of each network from resources controlled by the other.
  2. Two mechanisms, [exploitation and opportunity hoarding], cause durable inequality when their agents incorporate paired and unequal categories at crucial organizational boundaries.
  3. Two further mechanisms, [emulation and adaptation], reinforce the effectiveness of categorical distinctions.
  4. Local categorical distinctions gain strength and operate at lower cost when matched with widely available paired and unequal categories.
  5. When many organizations adopt the same categorical distinctions, those distinctions become more pervasive and decisive in social life at large.

A profound difficulty when durable inequality is present is that it is to some extent invisible (especially on the part of priviledged groups), since the unequal distribution adopts categories that are present in society at large. As a consequence, there may be some awareness that different groups are unequally represented in prominent positions, but if the mechanisms generating the unequal representation are not evident, then the explanation is sought in the nature of the individuals – for example, arguing that men simply are better at certain tasks.

I would argue that in contemporary market democracies, at least the four factors listed above are operative – markets, durable inequality, politics, and force. Only three of them are normally visible and subject to policy intervention. But one, durable inequality, is largely invisible (especially to priviledged groups), and is not easily influenced by policy. When envisioning possible future societies, it is important to bear in mind the mechanisms that give rise to categorical inequality, and ask whether they might be active and the impact they might have.

Inequality is Harmful

July 18th, 2005

Income inequality has been rising in the U.S. fairly steadily for about three decades, and the U.S. now features the highest level of inequality of any of the industrialized countries. There is some debate over whether this is a good thing or not. Some argue that a high level of inequality is beneficial, because it puts more money in the hands of people who will invest it, so higher inequality should lead to more rapid growth. While the evidence suggests otherwise (other things being equal, higher inequality is usually found to be correlated with lower subsequent growth, and the rise in inequality in the U.S. has been accompanied by a dramatic decline in savings rates), the idea continues to be influential.

However, even if the argument that higher inequality leads to higher growth were correct (and granting the implicit assumption that high growth is desirable in itself), high levels of inequality may be harmful. For example, one argument that inequality can be harmful is that there is less social cohesion in a relatively unequal society, because people see others as increasingly different than themselves as inequality rises.

In this post, an alternative mechanism is proposed though which rising inequality may have harmful effects. The argument that inequality fosters growth, and is therefore good for all, implies that aside from the beneficial effect of faster growth, rising inequality is neutral for those at the low end of the income ladder – the rising tide lifts all boats, even though the large boats rise more quickly than the small ones. However, this is not necessarily the case. My own observations have suggested that as inequality increases, it becomes increasingly difficult to find affordable housing if you are at the low end of the income scale. The reason why this might happen with increasing inequality is that housing costs tend to rise for everyone as average incomes rise, while the distribution of housing costs is narrower than the distribution of incomes.

Looking for evidence to test my thesis, I turned to the Bureau of Labor Statistics’ Consumer Expenditure Survey. Specifically, I used the historical standard tables, from 1984-2003, drawing data from the table Quintiles of Income Before Taxes. (, accessed 18 July 2005). Using the data from BLS, I calculated the ratio of expenditure on shelter to income after taxes for the highest-earning 20% and lowest-earning 20% of the sample population, and plotted it against the ratio of income after taxes for the highest-earning 20% over the lowest-earning 20%. The results are in the following graph.

Plot of housing expenditure share vs income ratio for top and bottom 20 percent of population

As can be seen in the graph, as income inequality increased, the highest-earning 20% of the population spent roughly the same share of its income on shelter. In contrast, the lowest-earning 20% increased its expenditure on shelter as a share of total income dramatically.

While there may be alternative explanations of these data, they certainly support my thesis and my own observations, leading me to conclude (as in my title) that inequality is harmful. It is harmful even if rising inequality contributed to increased growth (while the available evidence suggests it does not), because increasing inequality is not neutral for people at the bottom of the ladder. The BLS statistics suggest that in the past two decades, at least, the rising tide has swamped the smaller boats.

Relevance vs. Legitimacy in Quantitative Scenarios

May 30th, 2005

The overriding dilemma that faces a scenario team is to balance the demands of relevance and legitimacy, where by “relevance” I mean that the scenario analysis can make a meaningful contribution to people’s lives, and by “legitimacy” I mean that the analysis satisfies the expert community.

Balancing relevance and legitimacy is a dilemma that is particularly acute for a quantitative scenario analysis. Different stakeholders in the scenario development process place different weights on these two factors. For example, policymakers are mainly concerned with relevance – can they do anything with the information produced? At the same time, they are also concerned with legitimacy, in that it satisfies their own skepticism and provides them with leverage in policy negotiations. The scenario team itself is certainly interested in relevance, because they are interested in making a difference and in seeing their work acquire “legs,” but may be more interested in legitimacy, as this will satisfy the expectations of the team members’ profession and let them feel satisfied by their professionalism. In practice, meeting the goals of relevance and legitimacy are usually at odds. As the policy prescriptions get more interesting and range farther afield, the underlying analysis is likely to rely more on innovative techniques. In most cases, restricting the analysis only to well-developed techniques limits the scope of the scenario analysis severely.

Contemporary approaches to quantitative scenario development address the demand for relevance by focusing on a rapid development cycle that is responsive to stakeholders’ interests. How rapid and responsive the method actually is may vary with the method, the client and the practitioner, but the goal is rapidity and responsiveness. The challenge is to find effective ways for quantitative analysis to support this process in a timely, affordable, and legitimate manner.

Broadly, there are two approaches to meeting the challenge – by constructing a large, comprehensive model or by constructing a custom model for the particular exercise. In the first approach, the comprehensive model may have some flexibility, such as the ability to turn on or off portions of the model or to substitute one calculation method for another. (A good example of this is Barry Hughes’ IFs model.) In the second approach, the quantitative analysis team does not start from scratch. Instead, it uses common modelling patterns and a model database to rapidly build a custom model.

No single approach is best for all cases, and it is quite likely that no single approach will be best for most cases. Reaching the proper balance between relevance and legitimacy is a fundamental, persistent dilemma. Each study requires a unique resolution. The task of the scenario team is to make the conflict clear to the users of the scenarios, to negotiate the desired balance with the users whenever possible, and in all cases to make the position taken for the specific study very clear.