Let the magic do its trick
Rory Sutherland, zvg.

Let the magic do its trick

When politicians want to change the world, they focus on regulation and incentives. But in complex systems, more creative solutions are required.


Lesen Sie die deutsche Version hier.

If electric cars are more ecological, what is the best way to increase their circulation? If videoconferencing can make business processes much more efficient, why did we need a pandemic to start using them on a significant scale? When politicians and economists think about such questions, their ideas tend to revolve around two approaches: regulation and incentives. I have spent much of my life arguing that public policy leans too heavily on these twin levers. This is not to say that they are unnecessary, or necessarily wrong. But I do argue that the apparent certainty these two approaches provide comes at the loss of more creatively imaginative interventions which may be more effective and much less costly. Economics, by attempting to model itself on physics, creates no opportunities for magic: in this Newtonian model of the world the effects of any intervention are proportionate to its scale. In reality the world is complex, and in a complex world there is scope for magical interventions – for instance butterfly effects, where very small changes lead to massive change.

Psychology is a particularly ripe field for butterfly interventions: in my recent book “Alchemy”, I explain how a better understanding of human perception and emotion can obtain significant changes in behaviour. Many of these have already been tested, at times with astonishing results, usually under the label of Nudge Theory. In the UK, a simple experiment with “defaults” – in this case a company pension from which you are free to opt out, but where by default all employees contribute to the pension, have raised participation levels from single digits to around 90%.

But there is another area of scientific enquiry which also provides opportunities for magic: network theory. As in psychology, the same disproportionate outcomes may emerge from small, timely interventions in the right place. As with Nudge Theory, network theory simply requires that policy makers abandon Newtonian models of the world and embrace complexity.

It’s the network, stupid!

To demonstrate what this means, I shall start with a short language lesson.

A staple of upper middle-class dinner-party conversation in Britain is to bemoan the fact that the British, indeed native English speakers in general, are terrible at learning foreign languages. This has the advantage of subtly signalling to your audience that you yourself can speak a foreign language, so placing you among that sophisticated and cosmopolitan tribe which loves nothing more than to decry the narrow mindedness and stupidity of their less educated compatriots.

Yet by attributing this failure to laziness or stupidity, the speakers are revealing themselves to be rather dim themselves. They are making the lazy assumption that the reasons for this failure lie within the mindset or motivation of the people themselves, not in the wider complexities of the environment.

Anybody who makes the small effort required to think about the problem systemically – in network terms – will realise that there are multiple, unavoidable reasons why native speakers of English may be less than enthusiastic at learning foreign languages compared to everyone else.

  1. It is much more valuable for a Dutch speaker to learn English than vice versa.
  2. It is much easier for a Dutch speaker to learn English than vice versa
  3. If a native Dutch speaker starts learning English there is an immediate pay-off from day one: this is not true the other way around – indeed, there may be no pay-off at all.

What do I mean by “easier”? Well, if you are a native Dutch speaker (or a native speaker of German, Mandarin, Spanish, Arapaho or Tagalog) the first question “Should I learn a foreign language?” is usually easily answered. And if the answer is “Yes”, the second question, “Which language should I learn first?”, is a bit of a no-brainer. English is by far the most widely spoken language geographically. More important still, the fact that it is far and away the world’s most common second language means that its usage is not confined to any particular geographical area, nor is its value limited to conversing with people of any particular ethnicity or nationality. As one Swede explained to me, “We Scandinavians don’t learn English to speak to you – we learn it to speak to each other.”

Now, if you already speak English, it’s that second question which is the real hurdle. “Which language should I learn first?” And it is becoming harder to answer every year. Forty years ago, before affordable long-haul travel and the internet, it was simpler: “Learn the language of a large neighbour country.” Sure enough, until fairly recently, most educated Brits learned French, with some adding German or Spanish. Today, geographical proximity doesn’t really clinch it: Spanish, Portuguese, Arabic and Mandarin have competing claims, but are entirely useless in large parts of the world. At the age at which most people learn languages, they don’t know which parts of the world will be important to their future.

Remember, too, that – regardless of the grammatical and lexicographical differences between languages – it will always be much more difficult for an English person to learn Dutch than for a Dutch person to learn English. A Dutch person can learn English daily by reading subtitles in Dutch on English language television or going online. I don’t think I’ve ever seen a television programme in Dutch with English subtitles.

And for a Dutchman, having a smattering of English is useful, even if all you want to do is scan a US webpage or order a beer overseas. Then each subsequent improvement benefits you proportionately. That is not the case if an English person learns Dutch: here the benefits are neither immediate nor linearly delivered – indeed, there is a high threshold before it becomes useful at all. Apart from anything else, for the first few years of study, virtually every Dutch person you meet will still be far better at speaking English than you are at speaking Dutch. So why conduct a conversation in halting Dutch which your Dutch interlocutor could hold in fluent English?

Now, I am not saying the Brits couldn’t get a bit better at learning foreign languages. But it’s unfair to say British people don’t learn a foreign language because they’re lazy and stupid: the real root of the problem lies in network asymmetries.

Why cellphones spread faster than fax

And we need to study and understand these asymmetries better if we are to have any hope of understanding the world – and of improving it. If we want people to install solar panels, or to adopt wider use of videoconferencing, or to buy electric cars, it is hugely dangerous to assume that it is simply a question of individual will. The problem may lie in the network itself, not in the individual nodes.

The economist Douglas McWilliams writes extensively about this issue in his book “The Flat White Economy” – explaining why such network effects slow the adoption of many game-changing technologies, since it takes time for usage to reach the threshold of adoption where they become truly valuable. But for the most part it is an area of study ignored by most economists. And yet it matters a great deal.

I mentioned above the “scaling problems” and the “threshold problems” which are presented to any English speaker who wants to learn Dutch. These same problems also apply to the adoption of new technologies. There’s not much point in owning the world’s only fax machine, for instance, just as there is not much point in an Indian learning Breton.

Some technologies can naturally scale up at different stages of adoption. But what of those technologies – the fax machine being a case in point – whose value is only delivered at a certain scale of adoption?

The first fax machine was useless. But fax machines at least grew slowly because, in the very early days, a few people bought two. I had a friend who did this in the 1970s: he had two company offices, one in the US and one in London, and faxed information from one to the other. Only rarely back then did he use the machine to communicate with anyone else. But this was enough to sustain the technology (which was invented in the mid 19th Century) at a very slow rate of growth. When owning a fax machine finally became, with faster transmission speeds and the death of Telex, much more common, the desirability of the machines grew exponentially.

But it was still slow. Might it have paid telecommunications companies – had they a better understanding of network dynamics – to hand out free fax machines to certain select people? Would the gains from the acceleration in adoption have outweighed the cost of the free equipment?

Just imagine for a second how much more slowly mobile telephones might have grown had they only allowed you to call other mobile telephones. When it was first introduced, the cellphone was a niche product indeed. When I used one on London’s Oxford Street in the late 1980s people shouted abuse from passing cars. But the mobile telephone was helped by being to some extent bilingual – or in tech-speak interoperable: it could make and receive calls to and from landlines as well. In my early years of mobile phone ownership, I very rarely made a mobile to mobile call.

So one answer to the mystery of why videoconferencing as a technology was so slow to take off lies in network effects. In many ways it was closer to the fax machine than to the mobile phone. In order to replace a physical meeting with a videocall, you needed a large group of people to agree to shift the medium of communication from the default. Since that default behaviour was to hold a physical meeting, it was extremely rare to get a quorum to replace any proposed meeting with a remote version.

The free-rider opportunity

And this is why I think the COVID crisis has been so interesting in terms of its likely effect on videoconferencing. For, by removing, albeit temporarily, alternative forms of human meeting, it may have accelerated adoption of a useful new technology by 10 years. It was as though the entire world had spent four months when all communication was in Dutch.

The benefits have suddenly emerged because we were forced to reach the threshold of widespread adoption much sooner than would have happened had we left things to individual action. We have experienced a preview of a world in which we can commute less, work more effectively and waste less energy much sooner than would have happened if the sigmoid curve of adoption had played out in its traditional form.

Once this is all over, will we lose our ability to speak Dutch? I suspect we won’t.

Which raises an interesting question. I am not proposing that we manufacture annual pandemics in order to accelerate technology adoption. But are there ways in which the adoption of exponentially beneficial technologies might be accelerated by some other kind of intervention? What if the UK government had provided everyone with a Zoom account in 2018?

My friend, the Australian economist Nicholas Gruen, writes about this very wisely. He points out that, in most economic categories of good, there is a risk of what economists call “The free rider problem”: that people will seek to enjoy the benefits of public goods without paying towards them. But in networked goods, he argues, there may be a free-rider opportunity. Hence providing the poor or the rural with, say, subsidised high-speed broadband ultimately benefits everyone else far more than it would cost.

Network theory suggests that many significant effects might be achieved by short-term legislation deployed in a limited field – rather than through permanent statutes imposed on everyone. Or short-term incentives deployed simply to pump-prime a particular behaviour.

For instance, I don’t understand why governments didn’t recommend wearing masks earlier in the pandemic. Besides their direct effect, masks help psychologically as well as physiologically. Predictably, people were reluctant to wear masks later when it was recommended. Most people don’t have a problem with wearing a mask – they just don’t want to be the weirdo that wears one in the bus when everyone else doesn’t. As soon as the share of people wearing masks exceeds 10 or 15 percent, it will become common and even socially desirable to do so. And passing this threshold can easily be achieved in more creative ways than legal obligation.

Also, instead of subsidising electric cars, the state could help increase the number of charging locations. They could be made mandatory for new houses, and restaurants and clubs that offer reserved parking spaces for electric cars could have their costs covered. People are more likely to buy an electric car if they see their neighbours or guests at their favourite restaurant driving one, than because of financial incentives.

Purist economists may hate this. But if we are to intervene in a networked world, abandoning the dogma we have formed in a less connected world might be a necessary first step.

«Jeden Monat frische Denküberraschungen! Eine gehaltvolle und elegant gestaltete Zeitschrift.»
Francis Cheneval, Professor für politische Philosophie,
über den «Schweizer Monat»