On Jevons Paradox

The Jevons paradox is an observation that when progress or policy provides more efficient ways of using a resource for various purposes one expects the amount consumed (the demand) to decrease, but instead it increases.

Demand for coal increased after invention of the steam engine. “More car miles per gallon, more miles driven. Faster computer time, more time spent on computers.” (Quoting here from The ministry for the future.)

The explanation, in my view, is that control systems at the planning level, perceiving alternative means to an end, preferentially select more efficient over less efficient environmental paths. Consequently, when the means of employing a given ‘resource’ become more efficient, more people find ways of employing it to accomplish their diverse ends. Steam engines, computers, and vehicles with internal combustion engines thus came to be used for more and more diverse purposes, and that is why consumption of coal and gasoline increased.

Policy makers continue to assume that technological improvements in efficiency will reduce consumption. They fail to consider this consequence of the fact that the consumers are control systems capable of finding new uses for what they consume.

1 Like

This sounds to me like the Giffen Paradox that Bill Williams discussed. Is it different in some way? If it is the same, there should be some discussion of it in the archives from when Bill W was still with us. I forget his resolution of the Giffen paradox, but I expect some with better memories will be able to resurrect his resolution of the apparent paradox. I wouldn’t be surprised if is the same as the resolution you provide here, but I don’t have the energy to go look through that discussion in the archives.

Jevons paradox: as employment of a resource becomes more efficient (an increase in effect from a given amount of the resource), economics expects the demand for that resource to decrease (as though keeping the effect of use constant), but instead demand increases.

I have proposed that this is a collective effect of individuals finding additional ways to use the given resource. This is intended as an indication that we need further research into perceptions of alternative environmental feedback paths as we create, modify, and carry out plans.

Giffen paradox: As the cost of cheap foods rises poor people buy more cheap foods, violating an expectation in economics that an increase in cost results in a decrease in demand.

People eat foods of different cost and quality. Call them steak and bread.
Poor people can afford only limited amounts of steak in their budget.
When the price bread goes up, they can’t afford steak at all, and they eat more bread, even though it costs more.

Control of budget and caloric requirements has higher gain than does control of the preference for meat. Individuals control hunger/satiation in the relatively short term (more than once a day, usually), and for this, bread suffices if they can get enough. (They usually control health effects, but over a longer term, and without education or careful observation not obviously related to the steak/bread choice.)

No direct analogy is evident. Whether the ‘resource’ is money in the budget. or caloric intake (as a quantitative surrogate for hunger/satiation), or subjective preference for one or another good, or e.g. health benefit of one or another good, there is no change in its efficiency.

Bill Williams wrote this up in Chapter 25 of Hershberger (1989) Volitional Action: Conation and control. (I can’t locate a digital copy.) In March of 2004, Bill W. cautioned relying completely on his 1990 article in American Behavioral Scientist 34.1, a special issue devoted to Control Theory, because he disagreed with changes that the editor had made without consultation with the author.

Bill W. reviewed constraints and conditions which economists have devised to hem the Giffen effect in so narrow a definition as to justify excluding it from consideration. Some of these are included or implicit in the ‘necessary preconditions’ listed the Wikipedia article on the Giffen effect, as follows.

There are three necessary preconditions for this situation to arise:

the good in question must be an inferior good,
there must be a lack of close substitute goods
the goods must constitute a substantial percentage of the buyer’s income, but not such a substantial percentage of the buyer’s income that none of the associated normal goods are consumed.

If precondition #1 is changed to “The goods in question must be so inferior that the income effect is greater than the substitution effect” then this list defines necessary and sufficient conditions. The last condition is a condition on the buyer rather than the goods itself, and thus the phenomenon is also called a “Giffen behavior”.

In the 1989 book chapter, Bill W. says

the Giffen effect is the result of a relationship between a particular structure of preferences and a budget level. The preferences need not be concerned with physiological necessities, nor need the budget be either absolutely or relatively low.

In the CSGnet post cited above and here, Bill W. inveighed against the presumption that maximization is a fundamental principle in economics. It seems to me that optimization rather than maximization is the norm in nature and in PCT.

The principle of maximization assumes that preferences always take the form such that applying the principle of maximization will identify an optimum pattern of behavior-- and when applied in economics it doesn’t work. The maximization principle has been known not to work since Alfred Marshall introduced this problem into the mainstream of discussion in his Principles text in 1895.

He was responding specifically to a proposal that game theory provides a unifying framework for all behavioral sciences, defining a hierarchy within the lower levels of which PCT might find a place, citing (with some ‘erosion’ of the quotation) a paper that Herb Gintis republished in several later revisions:

The rational actor model assumes that agents have preferences reflecting their wants and the tradeoffs among these wants, and that agents maximize their utility by choosing from an action set that is limited by available information, material resources and time, cognitive capacity, and the agent’s physical capacities.

Herbert Gintis (2003) Towards a Unity of the Human Behavioral Sciences

Bill Williams provided background on the process of rediscovering this solution to the paradox, and its prior publication.

When I was a graduate student as a thesis I attempted to develop an alternative to the principle of maximization. I used the Giffen paradox as an anomalous case from which to make this attempt. But, I didn’t get anywhere, so I bluffed my way into a Ph.D. As some people will tell you I am good at bluffing. Thirteen years later, and still trying to find a method that would explain the Giffen behavior I encountered Bill Powers. He was interested in economics, but didn’t see a way to apply control theory to economic questions. Together we solved the paradox over a long weekend.

The difficulty that I had, knowing the economic side of the issue, and the difficulty Powers had knowing modeling, was that to solve the problem you needed to know more than any one single person at the time knew.

Knowing modeling alone wasn’t enough, and neither was knowing the economics. However, one the problem was solved by combining what we knew, I did a literature search and I found that the problem had been solved. See item 4 in the bibliography below. But, Beckman for some reason didn’t go on to apply control theory to enough other problems in economics to demonstrate its applicability. If you happen to read the paper which I wrote and Marken edited for “American Behavioral Scientist” be warned that some parts are all fucked up. Marken thought, as usual, that he was smarter, knew more, and etc etc, and he attempted to improve the paper I’d written with consulting me.


  1. Beckman, M. J. 1953. “Comparative Statics in Linear Programing and the Giffen Paradox.” Review of Economic Studies vol. 22-23. # 61.

Why call it a “collective effect”? Your explanation is at the individual level; it doesn’t depend on people interacting with each other for the explanation to work. It’s also kind of a dormative principle. You are explaining the increased demand by positing that there is increased demand (individuals finding additional ways to use a resource is just saying that people find new things to demand).

The PCT explanation is that control of food is the means of controlling both budget and caloric requirements. Expensive, more desired foods (like steak) are controlled for with higher gain than less expensive, less desired food (like bread). When the cost of an inexpensive food (like bread) goes up, people with a limited budget (poor people) have to increase their reference for the less desired, but now, more expensive food in order to stay within budget and maintain their control of caloric intake.

I had a java demo of this on the net but took it down when there was a change in the security requirements for java applets. I’m going to try to put it up again as a javascript applet since it is an excellent demonstration of the application of PCT in economics.

I think there is a general analogy in the sense that both can be shown to be non-paradoxical when looked at in terms of the variables people are controlling.

I was the editor and I don’t remember sneeking unwanted revisions into Bill’s paper. But I profusely apologized to him, on the net and in person at the Boston CSG meeting. But he never accepted my apology. And he never told me – or, at least, I don’t remember him telling me – what it was that he didn’t like about the published version of the paper. I just re-read his paper in American Behavioral Scientist and I can’t see what it was that he might have objected to. It is a very short paper and perhaps he thought I cut too much. But if you know what his complaint was, Bruce, please let me know.

Actually, that’s not quite correct. The control model produces the Giffen effect because there are “non-negotiable” reference specifications for the states for two higher-order variables, budget and caloric requirement. You can’t spend more than you have in your budget (no credit cards) and you can’t live on less than your caloric requirement. And the effect will not show up for people with budgets that are relatively large. It is a behavioral phenomenon that will only show up for people whose budgets are relatively small (poor people).

I think people do control for what could be called “maximization” by controlling for the first derivative of a variable, like money. I also think people, like engineers, can control for optimization – but in that case optimization is a controlled variable, and a pretty high level one at that. Optimization is certainly not the norm in PCT since control systems, even those controlling for an optimum engineering design, are controlling for matching a perception to a reference specification, regardless of whether that reference specification would be seen as optimum by an outside observer.

I wonder what those parts are. Maybe you could tell me what they are, Bruce. Clearly, I am not as smart as I think I am. And I really don’t think I am very smart. Don’t know much about history, don 't know much biology, don’t know much about a science book, don’t know much about the French I took. But I do I know 'bout PCT. And I know that if you knew it too, what a wonderful world this would be.

Because demand, in economics, is a quantitative measure of sales or the like in a population.

Collective control is always brought about by means of individuals in a population controlling their perceptions individually. However, it does not require the individuals to interact directly with each other. It only requires that what they do affects the state of the collectively controlled variable.

As an aside, one of the difficulties of influencing social change is that individuals can participate in collective control without even perceiving the collectively controlled variable as such, nor their influence on it and through it on others in the population.

No. Demand (see above) increases as a consequence of new employments for the resource being discovered and put into play by individuals in the population.

No, it shows up when the prices exceed the budget, and this can happen with any budget. In times of extreme inflation, or times of extreme scarcity, the wealthy also find foods stretching or exceeding their budget.

You could say that under those circumstances they become poor, relative to what they had been, but ‘poor’ is relative to others in a population. Anyway, that too is sophistry.

I am using the term in a generic sense that is older than the technical sense. Reminds me of Bill challenging my use of ‘dither’ in the ordinary sense that has been common usage for 400 years, and insisting on only the technical sense in computer engineering. (The context was description of an example of sequence control, where indecisiveness about the next step resulted in a gesture to pick up one object was interrupted to reach toward a second object, then back and forth briefly.

Just pretend that I said “It seems to me that sufficiency rather than maximization is the norm in nature and in PCT.” Sufficiency, adequacy, various words suggest approximation to the reference depending upon gain.

Maximization occurs in the idealized case of runaway conflict, which seems rare in nature.

Maximization is attributed to the ‘rational agent’ in the ideology called economics. IBill inveighed against this. It’s true that some people do seem to set a reference value at infinity for the accumulation of money. Because of its abstraction (fungible, can buy all kinds of stuff) it seems like more is better. Such a reference value tends to demand the available attention and means of control. As a teacher once told me, it’s easy to become rich; just think about nothing else. Pretty demanding. The proverb attributes the root of evil to the love of money.

I do not. It might be in his correspondence with Bill, which may be in the archive from Bill’s computer.

I’m sorry that you believe I don’t know anything about PCT, but it’s not my business to convince you that I do.

You’re right. I thought you said “collective control” but demand is, indeed, a collectively produced effect. It’s not really a controlled variable, though advertisers seem to think so. And they are pretty good at getting people to go out and demand the products they sell. So maybe demand is a controlled variable after all, controlled (somewhat loosely) by the advertisers and the producers who know how to produce what people want.

Ah, you’re right. Your explanation of increased demand for a resource with the development of more efficient ways to use it is S-R, not dormative principle: new employments for the resource (R) are a consequence of the development of more efficient ways to use the resource (S).

Well, then Bill Powers, who actually developed the control model that explains the Giffen paradox, is a sophist, just like me. He had this to say about it:

I think this solution of the Giffen Paradox has enormous implications concerning poverty, in fact providing a clear definition of what constitutes poverty. You are in poverty if your income is so low that when the price on the cheapest goods you buy is raised, you are forced to buy more of the cheapest goods and less of the more expensive and higher- quality goods. [Powers, CSGNet, 9/17/1993]

Sorry, I was just trying to make it scan with the song by the late, great Sam Cooke. I know that you do know something about PCT.

No, they’re a consequence of people controlling to find ways of controlling, or to find better ways of controlling. That’s what we do at the planning level. We try out different arrangements of sequences, testing in imagination whether what is controlled at the conclusion of one sequence is useful or necessary in the course of controlling a next sequence.

Factories already distributed watermill power through amazing systems of pulleys and belts, as we saw in Manchester. Substituting a steam engine enabled factories to be built in all kinds of locations apart from rivers, which could freeze or dwindle seasonally and were mostly not near cities and ports, complicating transportation. The invention of truly practicable steam engines by Watts was enabled by a partner who was a toy manufacturer, repurposing heuristics and mechanisms from that realm. And so on.

Yeah, I know. But words have consequences, starting with people’s perception that they are intended to mean what they seem to mean.

That’s a much better way to say it as a PCT hypothesis. It explicitly hypothesizes that producers are often controlling for building new factories in locations that may not be near rivers. The steam engine allows them to do that.

I am sorry; I shouldn’t have said it. But you’ve got to admit that your understanding of PCT is rather different than mine.

Rick, this comment is a bit late, but it seems to me that you choose to forget your role in this.

The special issue of ABS that you edited was published in Sept/Oct 1990. I remember you telling me when I came along in the early 90s that as editor you ran short of space, cut Bill Williams’ paper in half and inserted your own brief ending. All without telling him. This was discussed from time to time on CSGnet. You refused to apologize for ten years. When you finally found yourself pressured enough to “profusely apologized to him” it did not come across as sincere, did it? Now you play the victim. And you want Bruce to explain to you what was the matter. Seems pathetic to me.

And your point is?

I cut and edited almost every paper in that issue, some more than others. As I recall, the ones I cut most were those by Hershberger, McPhail & Tucker and Williams. The one’s I edited most heavily were those by Williams and Ford.

I spoke to nearly every author that I was going to cut and/ or edit, telling them that I was going to do that and asking whether they would accept my changes sight unseen. (I didn’t speak to Runkel or Powers because their papers were short and sweet).

And as I recall, every one, including Williams, said yes. I had to do it that way because I was on a deadline. As you may recall, the only way to get papers back to the authors for their approval at that time (~1990) was to send them by snail mail and wait for their return. I had no time to do this so I had to make the cuts and edits without giving the authors final review.

I don’t remember every call but I do remember one in particular. I called Frans in the Netherlands and asked if it was ok if I made some cuts and edits to his and Hettie’s paper without sending it to him for approval. I was glad that he said yes because his paper was great, well written but a bit too long if I was going to fit in all the papers. And I’m pretty sure I asked for the same approval from everyone I called and got it.

I apologized because I thought (and still think) it is possible (though unlikely) that I neglected to tell Bill Williams that I was going to make the cuts and the edits without his review. What more can I do than apologize?

And I’m not playing the victim; I clearly am the victim of your very dysfunctional discourse – the kind of discourse that you had so hypocritically disapproved of in another thread. This post of yours is just pure, unadulterated ad hominum hate mail. What am I supposed to do now? Fight back by telling everyone Bill Powers’ opinion of you (and your understanding of PCT) that he shared with me when he stayed with me in LA and I took him to visit you? But that would be so dysfunctional.

I can only assume that you do this(posting hate mail about me) because you want to discredit me as an expert in PCT. So I suggest that maybe it’s time to take Tim Carey’s advice and just let go of your hatred of me and start discussing what this group is purportedly about: PCT. You could start with an evaluation of this post from Eetu. I see some misconceptions in it about what constitutes a disturbance in PCT. I’d be interested in hearing what you think.