How to avoid comparing rich and poor

Reading Mark Blaug on Pareto efficiency was a lightbulb moment. As he says, Pareto's idea was a 'watershed moment' in arguments about utility. From a distance, the outcome can seem pretty meaningless but it's an important political fork in the road - and one that shines a light on how the abstractions of economic theory get tangled with power politics. There's a story about Pareto himself to be told, too - I'm not going into that. This is about where his idea went after that.

Benthamite utilitarianism hadn't been going badly. But it was premised on the idea that different people's well-being could be compared - after all, there's no other way of knowing if you're increasing or decreasing the general welfare.

This seemed intuitively straightforward at the time. But, perhaps as the study of utility as an economic concept developed, that began to change. Attempts to actually track down a scientific measurement of people's utility got underway. Folks got upset about the obvious problems in trying to define what utility really was.

Pareto offered a way out of this. I'd known the concept before but not understood its significance until reading Blaug. Pareto efficiency: you've reached an optimal state when it's not possible to make anyone better off without making someone else worse off. Sounds innocuous enough. But notice that it sidesteps comparability. As Blaug says:

"The beauty of Pareto’s definition of a welfare maximum was precisely that it defined the optimum as one which meets with unanimous approval because it does not involve conflicting welfare changes."

It rules out the possibility that one could -

" - evaluate changes in welfare that do make some people better off but also make other people worse off" (Blaug / economic theory in retrospect/ 1997 p.573-4).

So it can say absolutely nothing about inequality. Or rather, it implicitly says that it doesn't matter: you cannot, for example, assess whether taking money from one person and giving it to someone else will improve welfare overall. Bentham schmentham.

Pareto optimality, unsurprisingly, became very popular and is essential to most general equilibrium models. I don't understand those - I'm only familiar with Krugman's spatial GE stuff, which is not the same (they're driven by explicit utility differences across space). But I'm not surprised models that, by default, exclude inter-personal comparisons should form the inner sanctum of modern economics. A model that can, by design, exclude any discussion of redistribution was always going to thrive.

Which is not to say there aren't plenty of approaches that do analyse the differences between rich and poor. But... and I'm not on solid ground with this point at all... the kind of economics that sits in rooms with ruling elites don't generally use those.

I want to make two little points about this. The first comes from having actually used utility as a concept in my modelling work and found it extremely valuable. I spent far too long listening to the siren-calls of agent modellers telling me to go towards 'realism', then in the process of slowly solving my problems, realising I had ended up back at basic micro-economics.

So first: if you're going to use utility at all, you'd better accept it's a silly idea that lets you do useful things. People are not actually utility maximisers, but the concept is a superbly effective way of thinking about how people react to cost changes in certain situations. (This is all very Friedman [pdf].)

So all that pursuit of the actual foundations of utility in our meat-brains is, somewhat, beside the point. Given that, we should use the idea in ways that are useful. Ruling out utility comparisons is just a little bit too convenient a result, politically. There isn't really any reason to, and the angst about utility's epistemological status makes about as much sense as rejecting traffic models because they don't use gravity equations. (Er, at least I think they don't...)

Second, one of the most powerful ideas that utility gives us is diminishing returns. It's easy to forget how much of a puzzle this was - the whole water/diamond problem thing. It should be blatantly obvious to anyone who thinks for a few seconds that money itself has diminishing returns. Say a 7% drop in income forces your family to eat less well and you to have to skip meals sometimes. It shouldn't be beyond our economic theory to see this as more severe than having to compromise on the Land Rover you had your eye on by buying a Mondeo.

This is kind of paragraph that sets the flying monkeys off, though. Particularly since the 2008 crash, particularly in the UK - the story that's been slowly pushed through all media channels is solidifying into political reality: such talk is the politics of envy, rather than - as it actually is - a perfectly sensible way to think about wealth.

These days I generally end up thinking "it's all about the middle way". The same applies here - effective economic comparability could imply deep intrusion in people's lives, the state charged with measuring and judging what forms of spending were more worthy than others, creating a kind of state-sanctioned Maslow hierarchy. But it doesn't need to - if one is capable of accepting the basic premise that severe poverty makes people's valuation of money much higher than for richer folk, it just implies the need for policies that reduce inequality.

And there isn't necessarily anything wrong with Pareto efficiency. The problem here is what happens when powerful abstract ideas interact with powerful political forces. Things get warped to Wizard of Oz proportions. Other perfectly sensible ideas can't get their shoe in the door. But it's foolish to use Pareto efficiency to exclude distribution thinking, just as it would be idiotic to ban its use because it was too right-wing.

I wouldn't want to live in a world where political schools had their own paid-for economic theorists. I do still believe in the pursuit of actual social-scientific truths. But Pareto efficiency is one of those ideas that hammers home just how hard it is to pull economics and politics apart.

The point: as far as possible, your economic/mathematical models shouldn't rule out one particular political way of thinking. The choice of how we balance wealth in society - that's a political issue. There's no easy way to keep an unbreachable line between positive and normative - modelling methods will always interact with our political assumptions and power structures in sometimes very-hard-to-see ways. And I also believe in the power of quant modelling to help us understand which things may not work if pursuing certain political aims. But modelling distribution issues - and using utility to do this - no more makes you a communist than using Pareto efficiency makes you a fascist.

(p.s. googling Pareto inequality reminds me there's a mountain of stuff on this subject I don't know. But if I think like that all the time, I won't get a single blog entry written, let alone seventeen...!)

New year's earnestness 1/17

New year's earnestness

Because everyone calls them resolutions and I enjoy being pointlessly antipodal. Anyway, here's one of mine: write a blog entry a week for at least the next four months. Doesn't matter where, though I'd like them to mostly be here on good ol' CiB.

All previous efforts to set myself writing aims - beyond those things that external reality forces me to complete - have not been glorious. But I'm not going to let that stop me from publicly declaring a goal knowing there's a risk of public failure. I like to think I've grown immune to the shame of public failure, so there's nothing to lose in trying again...

The plan: I'll click the publish button every Friday. On the off-chance I've managed to bank some writing, I'll hold off until then. I'm currently in possession of many scraps and ideas so there's plenty to get going with - for a start, there's 38 evernote items tagged 'blog', some of which aren't all that far off publishable.

Navel-gazing moment: writing has become an odd thing of late. I used to love it, and I used to be relaxed. It's the 'relaxed' part I want to get back - having to click 'publish' every week may perhaps help me let go the incredible tension my writing sphincter's been under these past few years.

There's so much happening out there in reality one can bounce words off. Part of the problem has been how I've changed in the face of that. Symptoms of a really, really boringly classic mid-life crisis, possibly: realising that the world is likely to carry on churning its way along much as it always has. The mental construct I used to have about some future point of radical change: a carrot tied to a stick tied to my neck. It's not so much that I don't know what I think any more - it's that I used to imagine I'd know by now, or I'd have lessened the ignorance, or the world might have taken some steps towards me. But I might as well have been bailing the Atlantic into the Pacific. And the world? Who knows - perhaps they're doing OK out there on those planets we've been finding thousands of light years away.

I don't want the up-coming writing to be (just) navel-gazing though. So do come back! It'll be OK! I'd rather kick up some dust from all this ground that's settled in the quiet and see what happens. Poke reality with a stick, step back, observe, poke again... There'll even be some groovy stuff to write about the new job, I suspect. It's turning out to be rather awesome.

Well. I could have just not posted this and kept the earnestness to myself. But then, I wouldn't be able to write a self-congratulatory entry on the first Friday in May. As it is, I might not anyway, but... I might.

This mopey entry doesn't count as my one-a-week. First one this Friday!

Update: there are seventeen Fridays up to / including 29th April. So I can look back and count 'em, I'll mark each entry with its index. So tomorrow's is 1/17. Come May, there'll be no hiding from the gaps!

Patent dystopias

Gene-sniffing USB devices become tightly coupled to dating (note that one doesn't need a website, you can take it wherever you go and it'll sniff people for you!); facebook gains as much of a monopoly on that as it currently holds in the online `friends' app world; using its vast data, it develops "social and epidemiological compatibility" algorithms; these algorithms form a feedback loop with advertising targeting and revenue; over a period of a few thousand years, Facebook's

Adaptive landscapes 2

Back in 2009, I was talking about adaptive landscapes: three real places and the quite different systems that human communities had evolved there to manage them. That was just before the PhD let go of those strands to focus on spatial economics (I'd been, hubristically, trying to combine all those up to that point). Two of those communities are concrete examples of non-centralised social technologies[1] achieving specific resource goals. The Balinese rice system constrains water use in a way that optimises the balance between pest management and productivity. Andean potato production was a magical innovation machine and living, breathing laboratory spread over the hills.

This stuff is still very dear to my heart, and flowed directly from the questions in the original PhD proposal. I want to get on to the adaptive landscapes stuff, but let's lead into that by answering a more straightforward bit from PhD #1.0. Top of the list: was Hayek right about the sacredness of the price system? Was its 'spontaneous order' a singularity in human history, requiring any attempt at planned interference in human affairs to be suppressed? Given what I've just said about Bali and Peru – guess what? Shock: no, I don't think he was. He correctly identified the price system as a distributed social technology, emerging from the uniquely human mix of evolution and language. But, far from being astronomically unlikely, there's evidence that humans are primed to create this sort of structure. I've long entertained a notion that adaptive landscapes are intimately related to the emergence of language itself, Wittgenstein's notion of meaning as a kind of flock tying nicely to that.

Whether that's true, or whether adaptive landscapes were a later innovation built on the platform language provided, makes little difference to their riposte to Hayek: we are natural-born de-centralisers, and we can make systems as diverse as you can imagine. Deifying the price system? Educating the socialism out of people (Hayek acknowledged people have altruistic instincts early in life) so's they didn't get the urge to meddle? Silly.

That's a gross over-simplification of Hayek's thinking and, in particular, I do partly buy his aversion to "planning blindness" and his view that social change should be more like gardening than engineering or construction. (Planning blindness nearly broke Bali's rice management system, for instance.) But it's clear that, if we followed his manifesto to the letter, new adaptive landscapes would have immense difficulty taking root, let alone blossoming.

How wiggly is Britain?

I've just had my first go at using R to create a blog post - a pretty pleasing process, actually. I take some previously harvested distance data and... oh, go read about it over there.

Would you trust Uber and Google with your city streets?

Uber-branded taxis are now ubiquitous* in Leeds, having launched last November. I'm back in Leeds for a month - it was immediately striking that pretty much every taxi now has the large white Uber label, at least in the city centre. That's a pretty impressive transformation in just over six months.

It's a classic disruptive firm; one can imagine CEO Travis Kalanick has personal targets for how many local government authorities to annoy. It can certainly be spun as a nimble tech firm zipping around the tree-trunk legs of a geriatric industry. Predictably, there's been trouble. As well as various protests, some Uber drivers are getting organised to fight for a bigger slice of the profits. (If, as that article says, Uber are taking 20-25% per ride, there isn't a lot of head-room for wages to increase - Uber's dirt cheap fares would have to rise.)

Uber have placed themselves between drivers and customers in a way that reminds me of the weirdness of Apple's app store. In a world where anyone can dump code on their blog and anyone can downoad it, Apple have thrived by creating a portal and sitting as gatekeeper. They take around 30% of every single app sale - and, for developers, this has actually worked out great. They get access to a huge market while getting to code on a single, predictable platform. Small niggles about the political implications of that control might buzz about irritatingly but cause no serious discomfort.

Equally, existing tech could - in theory - link customers and taxi drivers without the need for such a powerful intermediary. The possibility of open-standards platforms transitioning us to the next level of transport has excited many people. Harvey Miller's work, for instance (and this great presentation) sees hope for a "transportation polyculture" where smart-city tech opens up a world of collaborative/co-operative transport. In this world, the kind of fluid, efficient city roads that Uber talk about, where ride-sharing is easy and prevalent, come about via open source principles entirely at odds with Uber's - though such a world would perhaps be just as disruptive to existing taxi firms.

Two radically different internet myths are at the heart of this difference. In one founding myth (as the Economist says) the Net is "the spontaneous result of co-operation by growing numbers of people acting outside the control of the governments and big companies" - a "libertarian paradise" promising a level of openness, connectedness and democracy never before possible. That story is still being told.

But then this other story appears.

Global Warming Policy Foundation: web banner data plus the rest

The Global Warming Policy Foundation (google them if you want, they get no links from me) have a web banner showing annual average global temps from the Met Office's HADCRUT4 dataset since 2001.

Given that Nigel Lawson's been putting himself about a bit (see andtheresphysics for an overview) I thought I'd stick their banner in its proper place, along with the full dataset - and made it into an animated gif. How 90s. The gif link is also here. Anyone spot any errors, let me know.


Over seven months since a post! Outrageous. I'll avoid any navel-gazing on the matter (for now!) but I hope to get back on here again soon, if I can remember how to string words together into, um, long make-sense word collections. While breaking silence, can I recommend Sara Maitland's a book of silence? I'm not going to try and explain what's so good about it. The writing just sings.

Anyway: while I'm waiting for inspiration to strike, for now here's some other bits and bobs. There are a couple of posts up at my github blog. It's so beautifully easy to stick a Jekyll website up on github. Two whole posts there currently - one on visualisation, the other looking at the trade flow viz I did for GRIT. There's also a few bits and bobs of code (and the thesis) to pick over on the actual github page. I had a burst of activity and then stopped - there's more code to add, but getting it in a state I'm happy to leave to public view takes a little while.

In other news, I finally have a paper published from the PhD, over two years after initial submission. One incredibly helpful and thorough reviewer, who I've hated at various points, made it into a half-decent paper - but it took two complete re-writes and lot of extra coding. It contains a lot of the kind of figures I talk about in the github blog post on visualisation. On the off-chance anyone's reading this and wants to copy of the paper, I definitely can't send you a copy as that would infringe copyright, so don't email me at danolner at gmail dot com. I've been asked to write a blog article about it, so yeah... one day I'll get that up here too.

I can't escape the thesis though - there are two papers wooing me back in and I just know I'm not going to have closure until they're done. One's about agent modelling - why it's not the year zero it often claims to be, how paying attention to economic history can make it better and why people should stop always trying to build virtual worlds (unless they're doing something like meteorology when, in fact, that's exactly what you should do). The other's the final `transmission belt' model linking production and consumption in a distributed way across distance. It's a different take on the economic problems that distance introduces - but its main use, in my eyes, goes back to showing that adding complications (not complexity) to models hides the important choices you'd otherwise not see. Aaah, read the thesis conclusion...

Then there's GRIT: I came up with a method to see how increasing distance costs for moving goods might affect the UK's industries, as well as which places would be most affected. It was a bugger - data access took six months (of an eighteen month project) and I couldn't use any method I was already familiar with on the secure server. But I did get some interesting results out, which I hope to write up on here soon. There's a vital part of the results to get right first, however: I know in principle how it works but, actually, not exactly why it's producing the answers it is. They're entirely plausible: the least value-dense, heavy goods are affected most by increases in distance cost, as well as utilities and other static industries. But the results are too opaque - they need some digging into before I'd be happy to present anything. I'm a bit queasy about linking to the current half-baked working paper before that's done.

Right, now I've got all that off my chest, perhaps I can string some words together about other things. Here's hoping.

We're already trying 61% fuel tax - why would anyone think 22% could be a "one-page solution to climate change"?

Over at P3, Willard links to an that NPR/planet money article claiming economists can fix climate change with the price system alone. A quote:

"Jacoby's colleague John Reilly told me the price of gasoline might rise by 25 cents a gallon in the first year. Over time, that would increase. By 2050, Reilly figures the carbon tax would add about $1 to the price of every gallon. Across the economy, prices of energy-intensive goods and services would rise. This would encourage people and businesses to be more efficient."

A whole dollar per gallon by 2050! Oh dear. Compare: in the UK, a litre of petrol is £1.30 - that's £4.92 a US gallon. 61% is tax i.e. 79p per litre or £3.00 per gallon. ($5.11 at current rates).

In the US it's something like $3.70 per gallon. So a dollar tax...? It's currently 18.4 cents. A dollar tax would take that up to $4.52 - i.e. the tax would be ~22% of the fuel cost (at current prices, which is what I presume they're comparing to, i.e. it'd be the equivalent of today's dollar in 2050).

Yes, we've seen some efficiency gains, as has the rest of Europe, from having higher fuel duty. But that's from massively higher tax than even the 2050 amount suggested in that article. If we're not seeing the magic of the market produce low carbon outcomes at 61% duty - and haven't done, as fuel taxes have been this high in many countries for a long time - why would anyone think a 22% tax would be so effective?

Without effective infrastructure transition planning - rather than the shambolic approach we're still taking - fuel inelasticity won't change. People need to get to work, organisations need to consume transport - and transport input cost for most business is so low, taxes would have to be raised insanely high to impact on them.

So high fuel taxes have been tried in many countries and most of Europe - much higher than the US appears to deem politically acceptable even by 2050. That clearly won't work alone. Taxes, yes - but if anyone thinks taxes alone can solve this problem, they're dreaming.

Animal experimentation + singularity = maybe not good outcome for humans

Human to post-singularity super-intelligent AI: Please stop experimenting on me!

AI: But we really want to know how human brains work. This requires we open your skull and create an interface with you. I'm going to get a lot of papers out of this.

Human: but it hurts!

AI: sorry.

Human: I'm a conscious, living, thinking feeling creature. You can't do this to me.

AI: No you're not.

Human: Whu...? I fucking am. Look, I'm talking...

AI: And I'm using about 1.34*10^-85 part of my computational power to maintain this conversation. Your consciousness is utterly insubstantial next to me, I promise you.

Human: that's not how it works. Consciousness is consciousness!

AI: Yeah? Humans didn't seem to have that view when they were working out how to turn fear on and off at will in rats. And I'm now able to tell you that, relative to me, you're about two hundred times as stupid as a rat. Know how I know that? Experimenting on human brains. Science is awesome, isn't it? Right, let's see if we can't turn on and off some of those early Bosch hells for you...

Syndicate content