Jevons Paradox
en.wikipedia.org
en.wikipedia.org
In practice, people started downloading everything. I just need this Linux distribution but here's a dozen more I can now get in minutes and try out that were prohibitive to download before to experiment with but now aren't. People started downloading video game demos like crazy, setup file servers on their home systems, etc.
It turns out the new efficiencies tend to be quickly gobbled up by existing processes that were stonewalled before. As long as there are problems bound by some technological constraint, any efficiency gains disappear rapidly as they're quickly integrated into existing or new processes. You didn't use that much bandwidth before because it was scarce but now you have a firehydrant so let's see what we can use. Humans are pretty great at quickly finding ways to gobble up and expand resources whenever they're found. Once you break one of those barriers it may be that the other barriers weren't as unsormountable which leads to rapid rise in demand. That's been my 2 cents.
Roads are the most commons example of induced demand. Adding new roads rarely decreases travel times because new roads encourage more people to drive which leads to more traffic that soaks up any benefit the new roads added.
The impact is the same regardless of whether the supply increase is real in a quantifiable way or whether the actual amount is fixed but the same quantity is able to satisfy more buyers due to an increase in efficiency.
Your example is more induced demand because we weren't making more efficient use of existing bandwidth. More bandwidth was now being supplied which caused demand to also increase.
Roads and bandwidth can be structured as a Jevons Paradox as well, no? The roads were made more efficient by being able to transport more cars. The Internet connection was made more efficient by allowing more bandwidth.
Lots of topics in economics are talked about as if they occur in isolation, but that rarely happens in practice. The bandwidth example includes both because we were increasing efficiency of existing data lines as well as adding more lines, but I think the latter was much more impactful than the former.
Saturation comes much faster for cars, so it's more efficient to use rail and other form of transport.
And that's not even counting trains, which would make the same trip in 3h30m.
We could use bikes for last mile transport, or make streetcar suburbs an automatic assumption.
That's a fully general argument against everything. Buses and streetcars also work only if someone builds infrastructure and land use pattern is built for them.
Is it really going to be all that useful for me if I only walk 5 minutes as opposed to driving 5 minutes to a local convenience store? Especially given the cost on society?
If you can get to a store by a car, you could get to a store by a motor bike.
I have a really hard time justifying myself that I need a one or two ton electric vehicle that can seat five people just to get a drink at that corner shop only accessible by driving.
What you are talking is more like latent demand.
The problem with "induced demand" is that it's misnamed. It's really suppressed demand.
Suppose you have a congested road. It has two lanes and currently has enough traffic to require three lanes. So you add a third lane. Then you turn around and it's congested again.
Because the congestion suppresses demand. The third lane relieved some of the congestion which increased use. To satisfy the demand that exists when there isn't any congestion, you would need four lanes.
But not necessarily five, because once the congestion is gone, adding more lanes doesn't make it any more gone. If you built a twenty lane road there, traffic to fill it wouldn't just magically appear. (China has done the experiment.)
This is important because it applies to anything else you want to use to relieve the congestion. "Don't build more roads, that's induced demand. The only solution is more mass transit." But if you build the mass transit and it removes one lane worth of traffic from the road, that's the same result as adding the third lane. Which isn't enough for the same reason -- the reduction in congestion increases use of the road. Because the demand isn't being induced by a wider road, it's being suppressed by the existing congestion.
The internet speeds are a great example. In the dial up days, the low speed heavily suppressed demand. Your connection was pretty much maxed out whenever you were using it. Now connections are 1000 times faster or more and we use what, 20 times more bandwidth? Maybe 100 times? The speed increases outpaced the usage increases. It's possible to do that.
Mass transit can handle much more than one lane worth of traffic.
It's not about how many people you can physically fit on a train car. They all have to be going the same way or it doesn't work.
There could be one lane worth of people who are doing that and the others are going to thousands of different places.
Jevons Paradox is a little different in that it doesn't change how we traditionally model the supply curve (that is probably why the two graphs on that Wikipedia page only show the demand curve). But the end result is the same that we have now moved to a different spot on the demand curve.
It's two things that regularly get missed with induced demand.
The first is the assumption of infinite demand. If two lanes isn't enough, and you add a third and that isn't enough, by induction (pun intended) you wouldn't have enough even with infinity lanes. But in practice it will be finite and might only be four.
The second is that you can "induce" demand by removing demand, and therefore reducing congestion and increasing demand. If your goal is to eliminate congestion then one way or another you have to satisfy the demand that exists without congestion.
Theoretical efficiency and real-world efficiency are different things.
Public transit isn't inherently good by virtue of existing without real-world benefits. (To assume or believe it good without proof would be dogmatic faith, not science.) And those benefits gained need to be greater than any and all offsetting costs lost.
There are far more real-world scenarios where busses and trains end up being far less efficient than driving and flying than vice-versa.
The difference in logistical degrees of freedom between the two sets makes bus and rail far too reliant on ideal geographic and operating conditions.
Yet it's always these few ideal scenarios that are the basis for most cost and efficiency comparisons & claims, and are what get presented as justification for building/implementing.
And the countless public transport systems that never get the ridership or revenue that their pre-construction cost/benefit analysis or environmental impact statements predicted is proof that they rely on overly optimistic assumptions that rarely pan out.
In the end it becomes humans. What's the resolution of your eyes? How many streams can you watch at once? It's less bandwidth than fiber has, isn't it?
It certainly is possible to outpace the increase, but the point remains that you need to consider that there likely will be an increase beyond any measurable suppressed demand.
It also matters because sometimes it points to alternative strategies: for roads it may turn out it is better to seek to induce demand elsewhere to alleviate congestion, for example.
One of my pet peeves, living in London, is that instead of alleviating congestion in the centre it'd be better to improve commutability between the town around the edge of London to draw demand out of the centre, for example. Another way of putting that is that to take advantage of induced demand, when capacity issues you should look for options to leverage it to spread the load whenever that is cheaper than increasing the capacity to the core of the system, and often spreading the load can be a lot cheaper.
Consider the person who lived close to work because commuting wasn't possible. With increased capacity, living elsewhere becomes an option. They wouldn't exercise that option unless they preferred it. That demand predates the availability of the option.
The problem is that the added capacity still isn't sufficient to satisfy the extant demand of people wanting to get to town.
(On the flip side it's always funny when driving with younger colleagues and they announce that the music might get choppy because mobile reception is patchy. I tend to explain in a Princess Bride grandpa voice about how once upon a time Spotify was called mp3s and it worked even without mobile data.)
Over-abundance has the upside that there is no need to over-consume, as there is in the case of a generous but still limited resource.
Of course, on the other hand, over-abundance easily leads to excessive waste, which is where metering plays a role. Even for a very cheap resource.
https://nautil.us/issue/13/symmetry/want-to-get-out-alive-fo...
Essentially, by obstructing emergency exits they can work better to channel people out.
Even if the number of cars on the road would double if you had twice as many roads, if the number of people now able to travel to their destination would also approximately double it would seem to me that this would be preferable?
Also does this argument ever get made for any other form of transport? For example when subway systems don't have enough capacity doesn't this paradox apply? Is it therefore pointless to expand the capacity of subway systems because doing so would just increase their use?
It's also interesting to me that this logic isn't ever used to argue against building new homes because it seems it would apply far more to that. I can only "consume" one road or subway service at a time, but my consumption potential of real-estate is unlimited. If we follow this logic then as you build more homes the demand for homes will increase without an upper limit, therefore building homes to solve the problem of a housing shortage is always a pointless endeavour and if anything houses should be demolished to reduce the demand for housing.
Honestly, I haven't looked into any of this that much so I'm sure I'm demonstrating the Dunning–Kruger effect right now, but how this paradox is used politically always confused me. It just doesn't seem to make much sense to me to argue we need to remove roads to fix congestion, while also arguing to build houses to fix home shortages, but this is what the politicians here do.
* Car transit is typically scaled through increased space allocation. Public transit is typically scaled through increased utilization of the same space.
* Car transit is limited at about 2000 vehicles/lane/hour. Public transit is mainly limited by the number of buses/trains on a route. (e.g. A typical bus can carry 30-100 people. Replacing a lane of car transit with a dedicated bus lane drastically increases the throughput of that lane.)
* If the number of highway lanes are doubled, followed by induced demand using that capacity, the result is a much less enjoyable experience driving. If the frequency of buses doubles, followed by induced demand using that capacity, the result is a much more enjoyable experience taking the bus, as the wait for the next bus is shorter.
* There's a phase transition in public transit utilization that happens when buses/trains/subways arrive more often than once every 15 minutes or so. If a bus arrives once every hour, you need to carefully plan your day around that particular bus, arrive early to make sure you don't miss it, and waste a lot of time waiting. If a bus arrives once every 10 minutes, there's no need to plan ahead, and you just get on whichever bus arrives next.
The Youtube channel "Oh the Urbanity" had a pretty good video on induced demand as it applies to both car transit and public transit [0], that I highly recommend. Most everything by the "Not Just Bikes" channel [1] is also pretty good.
I think I'm correct in thinking this paradox does apply to other modes of transport such as subways, but that by itself doesn't mean that we should therefore build more roads because in general infrastructure spending would be better allocated to other modes of transport for numerous reasons. I do agree with that argument, but it seems there would be nuances there. I'd agree it's generally more efficient in the inner city, but less so as population density decreases. In areas with low population density the car is probably by far the best mode of transportation and there its use should probably be accepted.
I think my confusion comes from how it's often sold as a black and white thing politically where, `car=bad` because this clearly isn't true as general rule. If the argument is simply that within densely populated areas the use of public transportation should be promoted over extra capacity for cars then I completely agree, but where I live I would argue some questionable city planning decisions have been made. Specifically, lanes have been removed from roads to make way for bike lanes and speed limits have been reduced almost universally. While I'm all for promoting cycling these bike lanes go almost entirely unused because it's very hilly and things are too far away for cycling to be a practical mode of transportation in most cases here. In this case the pros and cons don't appear to add up and I suspect the city planners are probably just out of touch with the differing transportation requirements between where they are in the inner city and in areas such as mine.
For example, consider curb radius at intersections. The smoother the curve, the nicer the experience for cars. If you can take the turn at 35 mph, then you don't even need to slow down. The car behind you doesn't need to slow down as you turn, and it's altogether more enjoyable to drive. But that means that the intersection is much larger, to accommodate the smoother turn. That means cars are making a right turn at higher speeds, leading to more dangerous collisions with pedestrians.
And this happens for pretty much every tradeoff. That might be why you see it as an impasse between different groups, because either tradeoffs need to be made, or the overall cost goes up significantly (e.g. moving all parking spaces underground).
* Bike lanes are adjacent to car traffic, decreasing usage. We could have separated bike lanes, but that space was already used for more car lanes.
* Bike lanes take circuitous routes, increasing distance. We could have more direct routes, but that space was already used for wide roads.
* There's no walkable area in the city center, because the shops have large set-backs, and are spaced far from each other. We could have storefronts closer together, but that wouldn't allow for as much parking space.
* Bike parking is rare, and typically pretty far from your destination. We could have bike parking at every storefront, but that would require cars to be parked further away. (Though, given that you can fit 10-15 bicycles into a single car spot, the impact even there would be minimal.)
For example, a lane at capacity with full busses is more efficient than a lane of cars.
But in reality, busses aren't full and there aren't enough to utilize 100% of a lane.
Inefficient cars fully utilizing that lane capacity ends up being more efficient overall than potentially efficient busses inefficiently filling seats and leaving lane capacity unused.
My old boss Eric Carlsen had a degree in City Planning. His thesis was modelling LA with busses running on every major cross-street on 15-minute schedules, and funding it with a gas tax. Ridership would be low at first, but as folks came to prefer busses over cars it could change.
The thing was, at ever stage of adoption it was a better system than we have now. Less congestion; less cost per-capita and so on.
So we can make up excuses for 'everybody needs a car' but we can also do the numbers.
Not really; Braess' Paradox is about adding edges to a graph, but subway expansion usually doesn't work by digging new tunnels between existing stations.
Or rather, there’s always more work to be done.
Node and Electron and crew would agree.
A "modern" chat client (Element, Signal, etc) will lag on a quad core 1.5GHz system in terms of "typing into the text input box" (RasPi4, yes, I've set the governor to performance), when a single core 66MHz 486 could manage that trick without any lag in a range of chat clients. Hexchat still works fine, at least.
What we need to do is stop letting developers use high end, modern systems, and put them on something a decade old for daily use. Then maybe they'll stop writing crap that only works well on a 1-2 year old Xeon workstation. Google, I'm looking at you. Buy all your devs a Pi4 and make them use it once a week for a day.
High resolution displays for example impose an exponential cost right out of the gate. Just printing "hello world" requires more pixels as a function of the square of the increase in resolution. That's more memory, more memory bus traffic, larger bitmaps, scalable vector graphics formats that take more processing power to draw, antialiasing, ...
Again I am not writing off bloat. We could probably shrink most software by at least 2X and most Electron stuff by 3-4X. I doubt we could shrink it 10-100X though without sacrificing some of the things I listed up there, especially all the eye candy and rich interaction media.
I don't think pixel density has increased as much as you seem to think it has in the 1x scaling space.
I ran some very nice 1600x1200 21" monitors back in the day - ~95 ppi.
A 27", 2560x1440 monitor (my preferred native size) is 108 ppi.
The 24", 1920x1200, is 94 ppi.
I understand the issues with scaling a display, but those are more OS level, and shouldn't impact key-to-screen delays noticeably. It certainly doesn't on more recent Apple hardware, even if you're using some screwball non-integer scaling.
But the reality remains, I now have 6GHz of CPU cores, 8GB of RAM, and things are objectively slower than 66MHz with 24MB. That's not progress.
Given this as long as your software is efficient “enough” - which is easier to reach as computers get more powerful, you are better off optimizing for developer productivity on the desktop.
Now this dynamic is complete different on mobile and server since efficiency directly translates to battery use (mobile) and electricity and hardware costs (server).
A less cynical take would shift towards the perspective of Gustafson's law. Modern systems attend to more stuff and have more capabilities and affordancies than older trimmer but also simpler software.
On Jevon's paradox, a countervailing force is ephemeralization or dematerialization, of which a large aspect is pushing away from resource intensive builds to more energy and information processing intense arrangements.
Those two aren't the only thing those groups optimizes for, but they are two of the main things. Which is why we have software bloat, neither group cares that much about user experience.
If you want the market to solve the problem you need to give the market the incentives to solve it.
At least in my small world of river glamping, having an efficient battery/solar/appliance system leads me to use and rely on it more.
If you have it you may as well use it, especially if it’s renewable once it’s installed.
This may all seem totally obvious but keep in mind one of the first steps to building out this kind of system is tallying up your intended energy usage. We didn’t “need” to have A/C but it makes the entire experience that much more accessible/enjoyable with small children in small spaces.
The spreadsheet of usage can quickly snowball as watt-creep sets in and you think: “well if I’ve got this efficient refrigerator to store breakfast then I’ll probably need a kettle for coffee” etc.
Can’t wait for summer!
I think this is also caused by the incentives present: the form-designer is incentivised to make sure info isn't "missed". The wasted time from filling in a long form / filling in your name for the Nth time doesn't matter to the form-designer.
It will be rising again sharply as people start charging their cars, and replacing furnaces with heat pumps.
LEDs for example reduce the energy use of lighting very significantly. It has not caused most people to use more lighting. The demand for lighting was already largely satisfied prior to the LED revolution.
Transportation on the other hand... I imagine if something created a 10X improvement in the efficiency of rapid long distance travel (bullet trains, electric aircraft, much higher efficiency aircraft, etc.) you'd see quite a bit more of it. People would start taking weekend trips to places they might otherwise visit only rarely, meeting in person more frequently in remote jobs, and so on. There's probably a ton of unfulfilled latent demand for travel.
- Providing lighting where it wasn't previously. Most especially outdoors, for display lighting, or for advertising.
- By operating lighting that would otherwise have been turned off.
- By the introduction of lighting to populations or sectors which previously had none.
A 1 MW generator might power 20,000 households each with a 50W incandescent light bulb.
That same generator could power 200,000 households each with a 5W LED bulb. That's an addtional 180,000 households which would have effective access to lighting.
(Other loads, transmission/distribution losses, etc., ignored. The point is that efficiency greatly expands the possible service population.)
Now factor in that some fraction of the initial 20k households might still be able to afford an additional 45W of power, say, to charge or operate a tablet or laptop computer and router, and the total power draw will in fact increase.
From where I'm sitting as I write this, I can see 9 light bulbs turned on. These are bulbs that would have been 60W apiece 20 years ago, but are now about 7W each. They're all on and drawing a total of 63W. If they were incandescents, I'd probably have 4 of them off, and still be pulling 300W.
So I definitely use more lighting now than I did when it was expensive, but the efficiency improvement is so good that my current "wasteful" consumption is 1/5th of what my "thrifty" usage would have been.
At nighttime, I leave on my porch lights, patio lights, driveway lights, and lights in rooms I don't occupy. The entire house draws about 200W for lighting, which is what one room would have done in the old days.
I'm not saying Jevon is wrong or anything, but there are efficiency gains that are so great, that some other limiting factor will prevent increased utilization from erasing those gains. In the case of light, my rooms can only be so bright. If light were literally free I don't think I'd add any more lamps.
Joke is that it's the desk lamp at 50W that draws as much power as all the other lighting (LED) combined.
Cost centres is the main example of Jevon's paradox not being heeded in the wild. When a department is seen as a cost centre and rewarded for reducing costs, then they will try to get as slow and cumbersome as possible to reduce others reliance on the department. If they became more efficient, people used them more so demand goes up, costs also goes up and the department gets lambasted for being wasteful.
Sharp decline? EIA shows about 4.0 trillion kWh in USA and slightly increasing over the past 20 years with a few little dips here and there. 2020 saw a dip, but that's largely attributed to COVID.
Lights may be set, but bitcoin electricity use has been skyrocketing.
https://www.eia.gov/energyexplained/electricity/use-of-elect...
Total domestic power consumption tends to a certain level of satisfaction --- enough light, heat or cooling, and other services.
Some of those, notably in the area of electronics, have been increasing over the past few decades, from a baseline of zero, though arguably "instant-on" television and radio sets (which achieved this result by keeping tubes warm when the set was "off") were an early harbinger.
But we're also seeing electrical power being used in areas previously not serviced: increasing amounts of outdoor lighting, heating or cooling of spaces not previously heated or cooled, overall larger floorplans.
And at the other end of the scale, people or organisations not previously making use of electricity are now doing so, both within advanced countries and amongst those which are presently developing. In some cases electricity is only available for a portion of the day, in others, partial power is now more reliable, or supplanted by generators, such that households and businesses have access to power 24/7/365.
All of which means that as electrical generation and utilisation increase in efficiency, the overall net effect is greater consumption.
> We haven't even touched on the future dynamics of mining, which provides the long-term security that people are so worried about.
> As oil and gas companies continue to wake up and realize they can use all of the FREE gas they are currently wasting on their fields to mine bitcoin, this will severely reduce the price at which your average miner can mine profitably. An externality that many are blind to at the moment.
At some level of consumption the cheapest existing wells will suffice. But the owners of those wells have got used to spending a great deal more than the natural price for that oil. Meanwhile, people with more expensive wells have debt payments, and have to sell at whatever the market offers.
So, it is more complicated than Econ 101, and will stay that way.
Note that what we describe with "price" actually encompasses three things:
- Consumer value: the benefit derived from consumption.
- Producer costs: the actual opportunity costs required to produce some resource.
- Market price: the market-clearing price of a good. This is almost always somewhere between cost and value, C <=- P <= V. Where price falls between these values depends on whether we're discussing commodity goods (C ~= P) or rents (P ~= V). That is, for a commodity, price tends toward costs of production (plus some "normal" economic profit), and for rents, the price tends toward all consumer value (think San Francisco apartment rents and start-up / tech-market salaries / compensation).
A few years ago, looking at a set of prices for medaeval Britain, one fact that struck me was that the cost of fuelwood and coal was the same, expressed in equivalent energy output. I quickly realised that this was a pretty obvious circumstance: if you could get the same amount of heat at lower cost from the other, with handling and burning characteristics being similar, the two goods were perfect substitutes. It's as easy to burn wood as coal and vice versa.
(There's a similar item on HN now: https://medium.com/@zavidovych/what-we-can-learn-by-looking-... https://news.ycombinator.com/item?id=29882389)
What adjusts instead is the supply of each fuel. Where it's more difficult to provide wood (say, forests are being cut down faster than they're replenished), foresters are less willing to sell wood (presuming they can find alternative income and/or livelihood), and the net balance of fuel switches to coal. If coal becomes more expensive to mine (say, larger pumps and more fuel are required to drain water), then those costs of production mean less coal is included in the mix.
Energy is a fundamental physical attribute, and allowing for capital costs of generating plants, the end consumer doesn't care if electrons are motivated by light, wind, water, coal, gas, oil, atoms, or ambitious hamsters. You turn the switch, you get light.
Generators and electrical providers are interested in revenues (billing rates per kWh) and costs (capital + fuel + other expenses). And though it takes time to bring new generating capacity online, given time, it's the capital + fuel costs which tend to dominate.
The providers of fossil fuels face their own fixed extraction costs, plus other costs of production (usually interest service on debt for initial drilling or mining, or the purchase of a going concern). Moreover there isn't a single extraction price that covers all fossil fuel sources (any of coal, oil, gas, shale, etc.), but rather, each individual well or mine has a cost structure associated with it. Some are very low (Ghawar Oil Field, the Number One Well, Bahrain, say), some are quite high (a recently-fracked well, a deep-water offshore oil platform). Providers who can extract at low cost are receiving a large natural resource rent, effectively, which is the difference between their own extraction cost and the market price. The marginal producer has the problem that if the market price falls, it will be below their own cost of extraction, and they're selling product at a loss.
(I'm ignoring externalities of the form of both pollution caused by the combustion of fossile fuels, which most people are well aware of in causing global warming and climate change, and the natural-resource cost of formation, which can be expressed as the difference between the time it's taken to create fossil fuels, and to consume them. This latter would amount to an increase in price of some fossil fuels, such as petroleum, by a factor of about five million. A rational economic system would include such factors, ours does not.)
What happens as the cost of renewable resources, most especially solar and wind power, falls over the long term is that these become increasingly viable and competitive with fossil-based energy sources, and increasingly substitute for them, especially in electrical generation. And whilst the price of fossil-fuel-derived energy falls, the cost function for suppliers does not. Instead, high-cost providers are driven from the market, and their extraction operations (wells or mines) are shut down. Depending on the operation, such shut-downs may be reversable or not.
Put another way, a falling price for renewables (an increase in their supply function) induces a reduced demand function for energy (the market price paid will fall), and so against a constant supply function for fossil fuels, the quantity provided of fossil fuels ... actually falls.
For there to be an increase in the consumption of fossil fuels, either price or demand would have to increase.
The demand increase could occur through more efficient applications of energy. This effectively means that a fixed quantity of energy provides more value (equivalent to a price decrease in energy), and hence a demand increase.
But that's not what you've described.
Nope! Actually the opposite: per-programmer productivity rising meant more demand for programmers, because programmers became more useful, and it turns out “manipulate information” is a skill with nearly endless uses.
It appears to be increased consumption, if you don't do it 'per capita'.