Way back in September last year, I concluded the post prior to Beyond this Brief Anomaly’s rather longer than expected hiatus by using a very simple example to illustrate the distinction between the way that causality is conventionally understood, and how it tends to be appreciated within systemic thinking. I’ll now round out that discussion by extending the ideas explored there to a real-world “problematical situation”, in order to show how our understanding of causality can have very practical implications for the ways that we organise things in the social sphere.

There are many examples that I could choose from that carry with them serious consequences measured in terms of concrete and uncontroversial financial, health, environmental or social indicators. These include development of antibiotic and pesticide resistance, arms races, and myriad other situations in which the pursuit of apparently straightforward technical improvements brings along with it co-effects—often but not necessarily unanticipated—that offset or undermine the expected benefits. These effects are often termed rebound, or in more extreme circumstances—where the co-effects result in an outcome that is actually worse than the original situation—backfire. Sometimes, the original initiative might achieve its desired effect, while also creating new problems of an apparently unrelated nature—one problem is solved while another is created. The kind of eventualities that can accompany an overly hubristic adherence to technological development—and that in doing so, reveal the consequences of too great a faith in the myth of progress—are illustrated rather poignantly by the story of the World Health Organisation parachuting 14,000 cats into Borneo in the 1950s, recounted in an earlier post. For present purposes though, I’ll zero in on a situation that has particular relevance in the context of the continued spread of automobile culture into the less-developed world, and that as such intersects in important ways with principal matters with which this inquiry is concerned. And as we’ll see a little later, this example has important parallels with—and hence lessons for—a pivotal issue for which this excursion into the structure of causality is preparing the path of inquiry.

Induced traffic

The situation that we’ll look at is associated with road construction as a response to forecasts of increased travel demand, and hence as an ostensible means of avoiding anticipated increases in traffic congestion. As with the examples mentioned above, it turns out that this commonplace and, at face value, unremarkable problem response is not nearly as straightforward as its proponents might have us believe. The perverse consequence—and by this I simply mean a consequence that acts to counter the claimed benefit of the primary activity—that accompanies road expansion is typically described as the induced traffic effect, or simply induced demand—although more correctly, this is a subcomponent of a broader phenomenon known as generated traffic. The phenomenon is characterised in the most straightforward terms as ‘roads cause traffic.’ Now, I’d like to acknowledge at this point that such a blunt summation of a complex socio-technical and economic phenomenon might ordinarily attract legitimate charges of hyperbolic over-simplification, even wilful ignorance of important nuance. In this case, however, my understanding is that the evidence supporting such a characterisation is widely regarded as quite overwhelming. This view has been long-held by many observers—according to Brian Ladd, in his 2008 book Autophobia, from as far back as the early twentieth century.[1] It has received strong corroboration from researchers including Goodwin [2] in the mid-1990s, and more recently by Duranton and Turner [3]. Duranton and Turner’s research is particularly significant, as it clearly identifies the effect over entire city areas, rather than at the scale of individual roads undergoing expansion.

The proponents of road building to reduce traffic congestion assume a characteristically linear view of causation in the relationship between road area and congestion. This view is presented in perhaps its simplest form in Figure 1. Under such a view, road building activity proceeds on the basis of a relatively straightforward cost-benefit assessment: if the financial cost of building a new road or expanding an existing road is less than the financial cost attributed to traffic congestion (e.g. the cost of lost labour productivity while commuting workers are sitting in their stationary cars), then a case can typically be mounted for going ahead with construction, pending the availability of finance.

Figure 1: Simple linear view of causation in the relationship between road provision and traffic congestion

The cyclical view of causation—again in perhaps its simplest form—takes the view presented in Figure 1 as given, and then also recognises that reduced congestion costs have flow-on implications for road use. If costs of congestion go down, then driving becomes more affordable, and road use increases through some combination of existing road-users traveling further and new road-users taking advantage of the available space. This has the effect of eroding the cost reduction benefit provided by the road expansion, increasing traffic congestion costs again (usually after some time delay during which the initial intervention seems to have been successful). This simple view of cyclical causation is depicted in Figure 2.

Figure 2: Simple cyclical view of causation between road provision and traffic congestion: the induced traffic effect

A significant contributor to the induced traffic effect derives from the observation that people tend to allocate a relatively consistent time budget to travel—there is a more-or-less fixed period of time that someone will be prepared to spend traveling each day. If travel time for existing activities reduces, many of us tend to willingly adapt our activities in ways that absorb the reduction, so that we use our full travel time budget again. For instance, we may be prepared to live further from workplaces in terms of distance, if travel time is not increased. This has clear implications for economically rational efforts to reduce congestion costs at the macro scale: at the micro scale of individuals deciding how to spend their time, we’re often prepared to shoulder costs that appear onerous to those interested in managing economies at a population scale. In short, the macro scale cost-benefit analysis fails to account adequately for why it is that individuals do what they do. At the micro scale of individual lives, we stubbornly insist on making decisions about how we’ll spend available resources on bases other than those codified in economic models.

As I’ve attempted to highlight, Figures 1 and 2 offer very simple representations of the distinction between linear and cyclical causation in the relationship between road building and traffic congestion. I’ve deliberately started at that level to allow scope for pointing out the constructed nature of any view of causality for this situation. To illustrate what I mean by this, I’ll now present a more nuanced view of the induced traffic effect, without taking into account any further significant factors beyond those included in the simple view. That is, I’m leaving the “problem boundary” unchanged—I’ll treat the situation itself as no more complex than that outlined above, but will show how the way that we understand what is going on can nonetheless be viewed in more complex ways that potentially reveal new insights. An important point here is to show how none of the views of causality presented are true or correct in an ultimate or absolute sense—rather, they represent more or less comprehensive ways of characterising what seems to be going on in the situation of interest to us, and are more or less adequate for appreciating and hence managing that situation. For instance, if we observe that traffic congestion cost improvements are eroded over time, but adhere to the linear view in Figure 1, the situation may well continue to confound our efforts to deal with it. Adopting a cyclical view may open up other avenues for dealing with the situation that would otherwise remain out of sight.

I’ll build up the more nuanced cyclical view in two steps, illustrated in Figures 3a and 3b. As will be obvious from Figure 3b, I’m constructing this more complex account by taking the single causal cycle or loop in Figure 2, and dividing it into two separate causal loops that can be treated as dealing with relatively independent aspects of the situation that nonetheless interact with one-another. The first loop—Figure 3a—in fact starts by taking Figure 1 and rendering it in cyclical rather than linear terms, by recognising that “high traffic congestion costs” and “reduced costs” are two states of the same variable, traffic congestion costs. Road building activity is carried out to reduce this variable, and in isolation of other considerations—i.e. ceteris paribus or “all else being equal”—such a response seems pretty reasonable (what is not reasonable, as it turns out, is the assumption, of all else being equal, itself). Figure 3a depicts a simple negative feedback loop—a causal cycle intended to reduce traffic congestion and the costs associated with it. The letters “S” and “O” in the diagram stand for “same” and “opposite” respectively. An “S” indicates that where the preceding variable increases or decreases, the succeeding variable will change in the same direction; and an “O” indicates the opposite of this—where the preceding variable changes, the succeeding variable changes in the opposite direction to it. If traffic congestion increases, then road travel costs increase also, and vice versa. On the other hand, if road area increases, then traffic congestion decreases (and vice versa). By following a specific change—say an increase in road building—around the loop, we find that road travel cost reduces i.e. it is indeed a negative feedback or balancing loop. Note that this still leads to a very simple depiction of the situation. It combines together two distinct types of variables—stocks and flows—without differentiating between them as rigorously as is needed to deal with all possible changes in the situation. A more sophisticated approach that does formally distinguish stocks from flows would be required for building a functional mathematical model of the situation. For instance, the variable road building represents a “flow” of new road area into the variable road area, a “stock” of all road area, existing and new. If road building activity reduces, road area actually continues to increase, but at a slower rate. For road area to decrease, we’d need to include a flow away from road area called something like “road destruction” (and as we’ll see a little further on, actually including this would not be so silly as it might sound at first, given what the cyclical view of causality reveals).

Figure 3a: A more complex cyclical view of causation in the induced traffic effect: a single negative feedback loop in which road building apparently reduces traffic congestion

A further aspect of the relationship between Figure 3a—considered in isolation—and Figure 1, from which it is loosely derived, is worth highlighting before we move on. While Figure 3a is notionally a balancing loop, i.e. it brings road travel costs back to an equilibrium state by correcting for upward excursions in those costs via negative feedback, the background narrative behind the linear view in Figure 1 tends to be that investing in new road area is inherently worthwhile on the grounds that it extends the benefits—rather than just reducing the costs—of road transport. In other words, it stems from an argument for the value of growth in road infrastructure. In general terms, this is the domain of positive rather than negative feedback. Within the circumscribed logic of the growth narrative, continuous expansion of roads delivers continuous increase in benefits (or at the very least, continuous reduction of costs).  As mentioned earlier, any particular road project will be justified on a cost-benefit basis—the financial benefits, including savings associated with reduced congestion, will need to outweigh the investment cost in order for the project to proceed. But at the aggregate level of road networks rather than individual roads, the logic from which I’ve derived Figure 3a—considered in isolation—is consistent with an apparently common-sense argument for continuous investment in new roads. The more road infrastructure we have, the less congestion we’ll have, and so in principle, building roads will deliver ever diminishing congestion costs (and ever increasing benefits of reduced congestion costs), up to the limit where congestion is effectively zero i.e. use of roads is not impeded to any degree by other road users.

That might well be the case in a world where “all other things remain equal”. In our case, for ceteris parabis to hold, road use would remain the same prior to and after road expansion i.e. road expansion would have no influence on road use—the amount and pattern of road use would depend entirely on factors other than the level of congestion and the costs associated with this. This, though, is a world other than the one in which we find ourselves: enter the second negative feedback loop depicted in Figure 3b. This loop also includes the variables traffic congestion and road travel costs, but now they influence the attractiveness of road travel and hence travel demand and distance traveled. That is, as road travel costs are reduced due to road building, road use increases (with some delay), increasing traffic congestion.

Figure 3b: A more complex view of the induced traffic effect, adding a second negative feedback loop to the single loop in Figure 3a

The two negative feedback loops, working in concert, act to maintain traffic congestion and the associated costs at a level acceptable to road users, regardless of the total stock of road infrastructure available: again, as Duranton and Turner put it, ‘roads cause traffic.’ If planners deem traffic congestion costs to be too high, and build roads in an attempt to reduce them, but individual road users are content to bear the high costs, then the benefits of road expansion will—in the absence of a very limited range of other possible co-interventions—be negated. In effect, reducing the cost of road use has actually increased the productivity of driving—road users can now do more for the same cost, and so rather than pocketing the saving (in time or money), they tend to spend the potential saving on more driving. Note, though, that the consequences of this are significantly more perverse than might at first be apparent from Figure 3b. When roads are expanded, at least a part of the additional vehicle distance traveled will typically be new road users. When congestion converges once again on its equilibrium level, there will be more users affected. While the cost per user may be the same as previously, the aggregate cost for all users is now higher. This higher cost shows up precisely where economic managers were hoping to reduce costs—at the macro-economic level. For individual drivers, there may be no perceived disadvantage, but for those interested in managing whole economies, an anticipated cost reduction has become an increased cost burden—an example not just of rebound, but of blowback.

And so we see that by constructing the view of causality in this situation differently—by adding further complexity to the original cyclic view—we’re able to tell the story of the situation in a way that reveals further nuance that might be of interest to those trying to deal with it. In the case of our particular situation, the options appear rather limited. Congestion charges may be the most expedient response—if individual road users are less sensitive to the costs of congestion than are economic managers, then one option that would work in favour of the economic managers’ goal within the depicted logic is to introduce congestion-based charges for road use. As congestion on a road increases, users are charged more to use it, and so there is an incentive at the individual level to avoid increases in road use that would increase congestion (or to avoid traveling at times of higher congestion).

One other option is also apparent though. A little earlier, I mentioned that we would need to introduce to Figure 3a a variable titled something like “road destruction” in order to provide a means of reducing the stock of available road area. Admittedly it would perhaps be a little less contentious—though I suspect not much so—to speak of road closure rather than destruction. In any case, the effect would be equivalent. This wasn’t just about making an academic point relating to stock and flow modelling though. As is clear from the logic depicted in Figure 3b and the discussion of aggregate costs above, one effective way of reducing overall congestion-related costs is to close roads, rather than to build new ones. Closing roads has the effect of making road use less attractive, reducing travel demand and actual distance traveled. Drivers either find other means of transport, or they reduce their travel demand (and the activity that it enables). Transport planners are increasingly recognising now that new public transport capacity must be introduced in parallel with road closures of some form if the changes are to result in a net shift away from car use.

Before wrapping up here, I’ll just note also the very handy segue that the example we’ve been looking at provides as I finally start to look in detail at the major topic of efficiency in the next post. The case for road expansion is often made in terms of various efficiencies. For instance, supposedly avoidable financial costs are portrayed as an impediment to maximising economic efficiency—if resource costs are reduced for a given economic output, efficiency of resource use increases. And energy efficiency improvement is itself often explicitly cited as a driver for reducing traffic congestion through road expansion. Vehicles sitting at a standstill in traffic with engines idling are consuming fuel that could otherwise be used for moving those vehicles. A compelling argument can be made for the value of keeping those vehicles on the move. But any initiative that does actually achieve this will almost certainly result in more vehicles using the available road space, even if this is a matter of, for instance, introducing new traffic controls rather than expanding road area. Increased fuel efficiency, then, does not lead to reduced fuel use (and reduction of the associated emissions)—in fact it may very well lead to increased fuel use and emissions. As I’ll discuss in detail as we proceed, efficiency improvements—at least as they tend to be conventionally pursued at present—are most effective for increasing the productivity of resource use rather than for driving overall reduction in resource use. This has profound implications for how we approach the energy-related dilemmas that we face as a global society.

References

[1] Ladd, Brian. (2008). Autophobia: Love and Hate in the Automotive Age. Chicago: University of Chicago Press.
[2] Goodwin, P. B. (1996). Empirical evidence on induced traffic. Transportation, 23(1), 35-54.
[3] Duranton, G., & Turner, M. A. (2011). The fundamental law of road congestion: Evidence from US cities. American Economic Review, 101(6), 2616-2652.