The Revenge of the Network


In 1700, a devastating earthquake struck the Pacific Northwest’s Cascadia fault, unleashing a tsunami which crossed the ocean and wiped out swathes of Eastern Japan. And nobody in the West knew about it for over two centuries. As Europeans expanded across North America, they picked up hints in native folklore about a tremendous natural disaster, as did later historians when they examined Japan’s prefectural records. It was not until the 1980’s, however, that geologists discovered the fault and were able to reconstruct what had happened, realising that the myths were actually reality.

Compare that with the Great Tohoku Earthquake of 2011 which led to the tsunami which devastated Fukushima. News of the earthquake travelled around the world in minutes, and the Pacific Tsunami Warning Centre issued an alert for all the areas bordering the ocean less than two hours later, allowing low lying areas to evacuate the vulnerable.

1700’s Japan was closed to the world so information about the disaster did not spread and it had no impact on anyone else. The 2011 event, however, led to Japan’s car and auto-component factories closing for 6 weeks, pressuring global supply chains, the oil price falling on expectations of reduced demand and a highly unusual G7 intervention in the currency markets to suppress the yen which was rising on expectations of repatriation flows.

The interval between the two events has seen the world wire itself up like a brain during childhood and adolescence as more and more parts have come into contact with each other and established links. If the world of 1700 resembled the mind of a toddler, with limited interconnectivity, it is now more like that of a mature adult, with all the different regions talking to each other. The planet three centuries ago had the same shape as it does now but because different parts of it had not yet established contact, they had little impact on each other’s functioning.

The brain is a particular form of network, a collection of entities (nodes) which are connected and communicate with each other. When one part of the network acts, information flows to the rest of it, potentially leading to action from other nodes. If you were walking down the street and saw a lion, the data received by your eyes would be analysed and recognised as a particular type of animal which would then trigger the part of your brain which recognises potential threats. At this point, another part of you would decide how to react and might send signals to yet another bit to put your legs in motion, while increasing your production of adrenalin, increasing your strength and endurance temporarily.

If we think of the world as a network, the past centuries have seen more nodes being added to it, and the connections between them enhanced. Japan, for example, was a node in 1700 – people knew it was there – but its policy of isolation meant that its connection to other nodes was very limited. The Pacific Northwest, by contrast, was not a node on the European network – it would only be in the latter part of the 18th century that it was “discovered”. Adding a further layer of complexity, several nodes that did exist have split up so whereas a policy maker nowadays might have to consider Israel, Lebanon and Syria, his predecessor only had to deal with the Ottoman Empire.

The number of connections in a network is given by the formula n(n-1)/2 where n is the number of nodes or more intuitively, by assuming that each node connects to every other node, but only once, so the link between a and b is not counted separately from the link between b and a. The result is that as the number of points on the network rises, the number of connections increases exponentially.

If we were to just think of states being the nodes, then the world would obviously be much more complex than that faced by our ancestors. However, if anything that under-estimates the scale of the problem. For countries are not unitary entities, they are themselves networks. While power is wielded by governments, policy is often influenced by other groups in society as information flows through the system. For example, the British government’s recent decision to grant visas to 5000 HGV drivers came as a response to pressure from industry groups that the post Brexit settlement needed to change to deal with the new facts on the ground. Proposed changes to planning laws were put on ice in response to information by another part of the network (southern property owners as expressed in the Chesham and Amersham by-election). Over time, the connections in the network will also change – business will be more connected to government under a right-wing government, unions under the left. The decline of heavy industry in the U.K. means that sector has less influence than it once did, while the City has much more.

Even seeing the world as a network of countries which are themselves networks slightly underestimates the complexity of the system, as globalisation has led some of the lower-level nodes to form networks with similar entities in other countries, which run in parallel with the upper, state-level network. The general reduction in interest rates seen during the Great Moderation led investors to seek yield wherever they could find it. The parallel growth in the Mortgage-Backed Securities sector in the U.S. provided a convenient outlet for such desires, leading exposure to the American housing sector to spread across the planet. When house prices started to fall in 2007, the losses, which in earlier times would have been confined to the States, rippled across the globe, causing systemic issues for the financial system in several countries. Governments were forced to act as trouble in one part of their local network, the financial system, started to infect the overall economy, but rather than contagion arriving through the top-level, state network, it had come through a parallel system of direct links between the financial nodes in each country’s network of which they had less awareness.

Rather than see the world simply as a network, we should therefore, see it as a network of countries where each country is a network itself, and some of the nodes of these country-level networks have formed their own parallel network with the equivalent nodes in other networks. It is a network of networks some parts of which form other, separate networks.

The overlapping nature of the networks means that events in one node now have multiple ways of impacting the others making it more difficult to understand the system. In 1700, a policymaker could not have known about events in Japan, so he did not need to know – if there was no way for the information to reach him, so there was no way for it to impact any of his responsibilities. In a simple network, where the only links are between states, the only way for events in the U.S. housing market to impact the U.K. financial system would be through a change of U.S. policy forcing the British government to act. A regulator would therefore need to understand the American system and its regulation. In a fully linked world, however, where contagion can spread directly from node to node, regulators  now need to have an additional in depth understanding of the connections in a different, lower-level network, as well as those at the higher, state level.

This is not however, confined to the financial sector. The British government recently organised an emergency loan to CF Industries, the country’s dominant supplier of carbon dioxide. The company had been forced to close its plants as the global spike in gas prices had rendered them uneconomical. While the issues in any particular sector might not seem hugely important, the government was forced to act when it emerged that CO2 is a vital component in the food sector, and the meat industry was days away from being forced to close. To predict the emergence of this crisis, the government would have needed to have an accurate view of the global gas markets (famously hard – natural gas is often called “the widow-maker” in hedge fund slang), an understanding of the carbon dioxide industry, a deep awareness of the meat packaging industry and the ability to pool that disparate knowledge. Having the knowledge in silos across government would not have helped, a policy response could only have been formulated had those doing so known about each separate issue.

The more connections in the system, not only do we need to know more about each node, but the more uncertain that knowledge becomes. Each node becomes less stable as it is exposed to a greater range of external events which can prompt it to change. The only real threat to butchers in 1900 was a lack of animals, perhaps due to disease. That obviously remains an issue, but it is now compounded by a potential lack of carbon dioxide and other chemicals. The more links, the more potential changes, and thus the more chance that our understanding of each node is outdated.

The implications of this for governance are profound. The more nodes in the system, and the more links between them, the more needs to be known to understand it while at the same time, the more likely it becomes that our knowledge is outdated. The more unstable each node of the network becomes, the more likely it is to require government intervention to fix. At the same time, however, the more connections there are between nodes, the more likely it is that actions taken in one area will cascade to another, creating unforeseen problems. The current environment is one in which a government is more likely to be forced to act by a surprise, but less likely to act with perfect knowledge of the situation, and less able to do so without causing problems elsewhere in the system.

Earlier in the year, some commentators argued that the recovery from the pandemic showed that man had finally gained mastery over the economy. Governments had, effectively, been able to turn it off, put it to sleep for several months and then re-start it without issue. However, the network had its revenge. While Britain’s supply chain woes, exacerbated first by a reliance on cheap foreign labour, and then its absence due to Brexit, have captured that attention of the media, logistics systems globally have been stressed by re-opening, with driver shortages across the continent and record queues at the Port of L.A. Gas price rises may be grabbing the attention in Europe, but several Chinese provinces have experienced blackouts, a phenomenon expected to last until the spring, due to rising coal prices which are also threatening the Indian power generation system. Governments across the globe are firefighting as re-opening teething troubles flow from sector to sector.

While governments are unique in the range of their responsibilities, they are not alone in the need to analyse the complex global system. International investors too need to understand the world to make decisions. While their track record is far from perfect, they do have significant advantages over politicians in a democratic system.

For a start, investors have a single purpose. They want to make money. Governments generally have several motivations. They probably want the economy to grow, but they may also want to be greener, they may want to be more independent of other nations, they may wish to re-distribute resources. Success in one policy dimension may cause problems on others. In many of these cases above, policy decisions taken for a particular reason (ending free movement in Brexit did restore “sovereignty”, phasing out nuclear power in Germany post Fukushima did reduce the risk of nuclear accidents) have created vulnerabilities which have been exposed when other parts of the system have been stressed. Had politicians had a greater understanding of the overall linkages of the system, they might have taken different decisions.

Secondly, investors have access to up-to-the-second, accurate information. Stock, bond and currency prices are updated in real time, allowed instantaneous calculation of the profit or loss on each position. At any point, investors can know how much they have made or lost. By contrast, the data government must deal with is often incomplete, as when Downing Street aides had to resort to their phones’ calculators to work out the number of Covid cases, sometimes wrong, as when Excel problems led the UK to under-count cases, and frequently out of date as in all economic data which is retrospective. As a result, it is far harder for governments to know the position on the ground at any point in time than it is for the investment community.

Most investors will also, as part of the capital raising process, commit to explicit rules over how much any position is allowed to lose before it is automatically closed. In combination with the above, this allows them to identify mistakes quickly and rectify them, or at least prevent them from getting worse. By contrast, it is far harder for a government to identify failing policies due to data issues, and there is no similar mechanism for exiting them. Indeed, given the glee with which U-turns are greeted by the press and opposition, governments are incentivised to persist in failing projects on the hope that something will turn up.

The ultimate sanction facing the investment community is that of the market. A period of prolonged underperformance for an individual will lead to the loss of a job, for a firm, it will lead to clients withdrawing their funds, and potentially the business failing. Over time, this will result in the less talented leaving the industry. Theoretically, this role is performed by the electorate in periodic elections. In the reality, however, the picture is more complicated for just as governments may have differing motivations for their policies, so citizens cast their votes for a variety of reasons, not just rendering judgement on the effectiveness of the government.

Politics is tribal. People vote so that their team retains power or, just as importantly, so that the other team does not get it. In Britain’s 2019 election, 14mn out of a total of 32mn votes were cast in constituencies which have returned an MP from the same party at every post war election. Given that both governing parties have been responsible for some large errors during that period (the ERM debacle for the Conservatives, the Winter of Discontent for Labour), the fact that this is so suggests that a decent portion of the electorate sees their job less as rendering judgement on the performance of the government, and more as expressing tribal loyalty.

We do not expect professional investors to get everything right, so, given their advantages over policy makers, we should expect even lower accuracy from governments. This is not an argument against democracy, the ballot box may not be used to assess performance in the same way that a redemption slip is, but it remains a partial method of judgement. Authoritarian systems share the same problems of complexity and incomplete knowledge that bedevil their democratic peers, but without the longer-term prospect that an electorate will eventually pass judgement on their failings.

If anything, it is an argument for more democracy, and more involvement by the citizens. A society where government had fewer responsibilities, would be one in which it took fewer actions, and made fewer mistakes. A government which did less, would be able to know more about a smaller range of areas, as it redeployed resources. A society in which citizens took more responsibility for the common realm would see fewer errors as more local knowledge would be deployed, and those errors would be less likely to seep into other sectors. A society in which citizens did not subcontract everything to the government, would perform better.

It may be that globalisation has passed its peak, and tensions with China, a desire to “on-shore” and the fragmentation of global alliances will lead the world to become less connected. But if it has not, and the trends of the past continue, then we may learn that the true price of an interconnected world is permanently lowered quality of governance.

Stewart Slater works in Finance. He has been published by Areo and Spiked, and writes (when he can think of something to say) on Medium as Stewartslateruk