AI for international economists: Network effects, part 4 of 5
VoxEU Blog/Review Productivity and Innovation

AI for international economists: Network effects, part 4 of 5

In the 4th post in his series on AI, Richard Baldwin turns his attention to network effects and Metcalfe's Law, which helps account for the wild stock market values of companies that had never been heard of a couple of years ago.

"If you want to go fast, go alone. If you want to go far, go with others." Or so claims an old saying. The laws covered in previous posts were about speed. They are the reason why the digitech revolution is so swift and profound. This post turns to networks. 

Moore’s Law and Gilder’s Law are all about capacities, Metcalfe’s Law and Varian’s Law are different. They explain, describe and predict the transformational economic and social impact of faster processing and transmission. They show how people and companies turn capacity into useful products and services. Metcalfe deals with networks, and Varian with innovation. 

Metcalfe’s Law

How valuable is a network like the telephone network, the internet, or Facebook? Questions like this led Robert Metcalfe to formulate his law. 

It is obvious that networks are more useful, and useful more often, when they link more people, more computers, and more information. But this is not where his insight lies. Metcalfe’s Law states the value of a network grows twice as fast as the number of people connected to it. (Technical note: It asserts that the value of a communication network is proportional to the square of its size, since the number of potential connections with ‘n’ users in n(n-1)/2.) 

We can illustrate his logic using the story of how email caught on in the 1990s. We could tell a similar story for WhatsApp, SnapChat, and any other social media platform. 

Until around 1995, internet email was effectively funded by the US public, and the government explicitly discouraged commercial activity. When the public funding declined, private providers like America OnLine (AOL) took over and created ways for millions to have access to email. The uptake was slow at first. 

Not everyone had an email address, and many people who had one refused to use it. You could not, as today, assume that the receiver would read or even receive your email. Becoming an email user was a conscious choice. Some technophiles jumped in. Many others, like me, were unconvinced. 

The key question was whether the hassle of using email would be compensated by the benefit of making fewer phone calls and writing fewer airmail letters. This, in turn, depended on how many other people were using email. In other words, the value of the network depended on its size. 

At first, I used email mostly at work. What really mattered to me was how many other economists were using it. The number grew gradually at first but, by the end of the 1990s, it had become the standard method of communication for economists around the world. Once everyone else communicated by email, you had to do the same – not least because they all assumed you were using it and so were not sending invites via post or fax. 

Interestingly, one of the first economists to dwell on technological progress was a holdout. In the early days of VoxEU, I invited Nobelist Bob Solow to write a column and he replied – in an airmail letter – that he was watching the whole internet thing carefully, but wasn’t ready to participate just yet. That was 2007.

Tipping-point economics

Almost anything that involves cooperation will trigger the forces described by Metcalfe’s Law. This might mean new communications networks. So today, some people have abandoned email. Most young people greatly prefer communication tools like WhatsApp or Slack. My children, for example, consider email to be an ancient technology that is mostly useful for sending files to old people like me and their professors. 

This is why Metcalfe’s Law applies to many things beyond the Internet and email. Examples include consensus on certain file types like PDF and JPEG, or using programmes like Microsoft Word and Excel. Cryptocurrencies cannot escape the logic of Metcalfe’s Law – bitcoins, or any means of payment, get more useful, and therefore more likely to be used, the more people use them. 

Metcalfe’s Law explains why tech is wild

Metcalfe plus Moore plus Gilder produces things that sound insane – like Facebook being worth ten times more than General Motors, even though GM makes 10 million cars and trucks a year while Facebook is nothing more than a very large digital bulletin board. Facebook was launched in 2004 but was only opened to the public in September 2006 (initially it was only for university students). Five years later it had 600 million users. In 2017, it welcomed its 2 billionth user, and is worth a half trillion dollars. That is growth that would not have been possible if Facebook was limited by the laws of physics that applies to goods. 

Explosive growth in processing and transmission speeds make it cheaper and easier to connect to networks like Facebook or Snapchat. Since the value of these networks grows twice as fast as the membership, you can see why the economy in cyberspace acts so differently to the economy in meatspace. It accounts for the wild stock market values of companies you had never heard of a couple of years ago. It also has important implications for how digitech transformation of globalisation and robotics – what I call globotics in my forthcoming book, The Globotics Upheaval: Globalisation, Robotics and the Future of Work.

Metcalfe’s Law also helps explain why the virtual economy is often a winner-take-all contest. 

Winner-take-all and what it means for the pace of the digital transformation

In the 2000s there were competitors to Facebook. News Corporation paid $580 million for one of them, called MySpace in 2005. But everyone wanted to be on Facebook since everyone else was on Facebook, that was where you could find your friends. Metcalf’s Law in action is why News Corp ended up selling Myspace in 2011 for $35 million. 

Metcalfe’s Law will be a key driver of the automation of the service sector, which will result in it opening up to global wage competition. This is because we often don’t do something because other people don’t do it. But when others start doing it, everyone joins in. That is the basic logic of Metcalfe’s Law. Or as economists call it, network externalities. 

This will be critical when thinking about what I call telemigration (hiring remote office workers who are based in low-wage locations). This is already happening in web development, yet many companies in the US and Europe are unaware of the possibilities, or are hesitant to explore them. But that can change quickly. 

As the historian Ian Morris observed in his fantastic 2010 book, Why the West Rules—For Now: The Patterns of History and What They Reveal About the Future, “change is caused by lazy, greedy, frightened people looking for easier, more profitable, and safer ways of doing things”. 

When it comes to digitech, today’s corporate world has progress from sloth to greed, and more recently from greed to fear. Something like one in six companies have been dropped from the S&P500 stock index – shoved aside by innovative upstarts, many of which are embracing digitech solutions or products. 

Once a firm’s competitors start using low-cost foreign labour though, competition between these firms will accelerate the process. As more companies in rich nations source labour this way, more talented foreigners will join the network platforms looking for work, and more companies will join the platforms looking for remote workers. 

It’s coming faster than you expect

In the technology business, these matching platforms can grow very quickly. Take WhatsApp – people started joining once other people they knew started using it to send messages. When there was a larger audience, this convinced more people to choose WhatsApp to send their messages. In the 16 months to July 2017, half a billion people used WhatsApp for the first time.

My guess is that tipping-point economics will define the progress of global outsourcing of services. Once people begin to trust telemigrants – as they have rapidly come to trust consumer services like Airbnb – this sector may become very big, very fast. And raging demand will be met with raging supply. There are millions of workers in developing nations who would like to provide those services. As the numbers of buyers and suppliers rise, the network becomes more attractive to new buyers and suppliers.

It has already started – the world’s leading web platform for international freelancing,, was valued at $1.5 billion at its IPO on 3 October 2018. Shares rose 40% on the first day of trading.

Robert Metcalfe the man

Born in a distinctly non-digital world in 1946, Robert Metcalfe realised that the drive to digital would shape the future. After earning two Bachelor’s degrees at MIT (in Electrical Engineering and Management), he shifted two subway stops to Harvard, where he added a Master’s in Applied Mathematics in 1970. While working on his PhD in Computer Science at Harvard, Metcalfe took a job back at MIT that exposed him to the nascent internet – the then experimental project known as ARPAnet. The ‘ARPA’ stood for Advanced Research Projects Agency, a research-promoting arm of the US Department of Defence. 

ARPAnet linked computers at Pentagon-funded research institutions, using telephone lines. I used ARPAnet to pose questions about Fortran coding to my brother in 1975. I was doing economics at the University of Wisconsin-Madison; he was doing computer science and cryptography at the Xerox Palo Alto Research Center. Long-distance telephone calls cost many dollars per minute, and were beyond the reach of most student wallets, so ARPAnet allowed many of us to have conversations that would not have otherwise happened. 

ARPAnet’s revolutionary potential was immediately apparent to Metcalfe. He decided to write his PhD thesis on it. He moved to Xerox PARC – Xerox at the time was one of the most ambitious tech companies in the world – and it was there that he made his first contribution to the digital world. 

In 1973, Metcalfe invented ethernet, the key engineering design that still ensures that computers of all makes and models can be networked together, and can connect to peripherals like printers, routers, and external hard drives. On 22 May 1973, he circulated a typewritten memo, Alto Ethernet, including this schematic sketch. David Boggs, Metcalfe, with others, got a version of it working by November 1978. The patent application, filed in 1975 by Xerox lists Robert Metcalfe, David Boggs, Chuck Thacker, and Butler Lampson as the inventors of ethernet.

It is not hard to see how Metcalfe, having invented the linchpin technology that allowed computers to join a network, came up with Metcalfe’s Law about the value of such networks and how it is related to the number of computers attached. In 1994, George Gilder credited Metcalfe's Law as one of the foundations of his law.

Metcalfe founded 3Com, a manufacturer of computer networking equipment, in 1979. It took off after he convinced Xerox, minicomputer maker Digital Equipment Corp (PCs were launched only in 1981), and chipmaker Intel to adopt the ethernet as a public standard, and he became rich. He is now a venture capitalist. 

The limits of Metcalfe’s Law

One of the strong assumptions behind Metcalfe’s Law is that all the possible connections are equally valuable. In most cases, that’s just not true. 

Think about the telephone network. You call only a dozen numbers regularly and fewer than 100 in a year. There are 4 billion phone numbers you could call, but the probability that most of those will be valuable to you is vanishingly small. When the number of telephone subscribers rises from 4 to 5 billion, the telephone does not become more useful to you. Likewise, the average Facebook user has only about 150 ‘friends’, but the network as a whole still gets more valuable to more people as it grows.  

This observation doesn’t repeal Metcalfe’s Law, but it means that it should be taken as a rule of thumb. For most of us, the value of the network peaks fairly quickly. The dominance of really big networks comes from overlaps between a large number of moderate-sized networks. This limitation is also one reason that the internet is fragmenting. Many experts expect that in just a few years, there could be two or three different internets in the world.

Language and regulation are also huge barriers to the spread of networks. For example, in China, Facebook, SnapChat, or WhatsApp are not popular. Instead, platforms like QQ, QZone, WeChat, and Sina Weibo have millions of users. Part of this is driven by government intervention, but it is also the segmentation of users that language and distance naturally create. Culture and language still matter.

Metcalfe’s Law links the number of connections to the value of the network. Varian's Law, the subject of the next post, is related to this, but deals with the implications for innovation.

Read the next blog in the series here.

Read the previous blog in the series here.