Network Risk: The Unseen Aspect of Portfolio Risk

At the start of this year, Carillion, one of the UK’s largest construction and engineering firms, filed for compulsory liquidation. Citing mounting debt as the primary driver of this action, the immediate focus was on the 43,000 jobs worldwide—all direct employees of the company—that would be lost as a result.

In the days following the announcement, both the financial and mainstream media were filled with stories on the wider impact of Carillion’s collapse. Ranging from interviews with CEO’s of small listed firms to unlisted owner-operator outfits and independent contractors, the interconnection between securities that provided supplies and services to the construction giant began to emerge. On January 21, the Financial Times published an article highlighting that Carillion used 30,000 sub-contractors, many of which were expected to fail as a direct result of this liquidation.

These widespread effects implied some common driver of volatility among Carillion’s suppliers; however, in the immediate aftermath the specifics were difficult to determine. An easy assumption would be to cite the Construction and Engineering industry as the major driver, but that would fail to account for the impact on companies such as Gattaca PLC (an AIM All-Share listed engineering & technology recruitment solutions company) which was forced to issue a statement regarding its price decline in the days after Carillion’s liquidation.

In fact, looking at the wider markets, on the day of the announcement the FTSE 100 was set to rally while the FTSE All-Share Industrials index, of which Carillion was a member, barely registered the announcement:

Breaking Down the Network Effect

In the absence of a common market driver, there are two potential explanations for the impact of Carillion’s collapse on the wider network of companies. The first possibility is that this was an idiosyncratic event for every company in the Carillion network. However, the commonality between these suppliers and the effect of Carillion failing is contrary to the nature of an idiosyncratic event. The second explanation is that this was a market shock, implying that Carillion was a systemically important organization, such that its decline would force the collapse of wider network of companies. While this appears to be the case, it is difficult to understand how the failure of a company that comprised less than 1 basis point of the FTSE All Share and just 3 basis points of the Industrials sub-index at the end of 2017 could have such wide-reaching implications.

More important, however, is an understanding of how companies such as Gattaca PLC—which no commercially available risk model identifies as having an exposure to the Construction & Engineering industry—could be so heavily impacted by events in those markets.

Could the answer lie in understanding the sensitivities between a stock and its supply chain? When combined, these datasets provide deeper insight into the network of companies to which a given stock is exposed. If subsequently combined with the power of an ex-ante risk model, could this provide us with an understanding of the wider risk factors that a company is exposed to above and beyond its own direct risk sensitivities?

Defining the Network

To address these questions, the first step involves identifying the network of companies for a given stock. Using FactSet’s Supply Chain data, we can identify the different relationships—suppliers, customers and partners – that a company has. Proprietary methodology ensures that relationships are identified through not only direct and mutual disclosure, but also reverse disclosure ensuring the most accurate picture of a company’s network is captured.

With this detailed supply chain data, we can define a network of stocks to which a given security is connected, comprising either customers or suppliers or a combination of the two. With the network defined, each security can be represented as a portfolio of stocks, where the stock portfolio includes both the original security and all its customers or suppliers.

Taking BP PLC as a well-known, globally significant example, this approach allows us to go from a single stock which is characterized as having 100% exposure to both the UK and the Oil & Gas industry in the first image below, to a portfolio of over 260 core suppliers having both global representation and less than 40% weight attributed to the Oil and Gas industry in the second image.

Using this relatively simple example, it’s clear how a company that is classified as a having a membership to a single country and industry could be impacted by market events in other countries and industries. Looking back at our Carillion example, it becomes more obvious how Gattaca PLC could be impacted by the collapse of a customer in a different industry.

Moving Beyond Single Stock Analytics

Using the same approach, we can aggregate this information at the portfolio level. Taking the FTSE 100 as an example, we can expand the 101 constituents into separate network portfolios consisting of over 1300 customers and more than 3500 suppliers:

This now reveals the FTSE 100 to be much more than the UK-defined benchmark:

Viewed through the geographic deployment of its customers, we see that the benchmark holds 24% in U.S. equity:

From a supplier-based perspective, the benchmark becomes almost 30% comprised of U.S. equity and almost 5% Australia equity:

Similar analysis can be carried out against the DNA of our expanded FTSE 100 index. By measuring the sensitivities and tilts of the characteristics of the expanded supplier-network FTSE 100 against the core FTSE 100 index, we’re able to see that the benchmark holds a tilt towards small cap, low-quality, volatile stocks:

From the customer base, we see that the benchmark further includes a growth bias:

A Hidden Source of Risk

This relatively common “look-though” analysis, often used in fund-of-fund or multi-manager mandates, can provide a wealth of insight into the underlying characteristics of a fund that, while easy to understand, can be difficult to illustrate. Going back to our Gattaca PLC example, it seems intuitively correct that a firm which is involved in recruitment for the Construction and Engineering industry would be impacted by the collapse of a large company like Carillion. However, without insight into Gattaca’s customer base, this is more difficult to show.

Taking this idea one step further by incorporating an ex-ante risk model into our analysis, we can provide additional insight into potentially hidden sources of risk and return.

Here, we see that the base FTSE 100 index has a tilt towards large cap and income and a tilt away from medium-term momentum, as defined by Axioma’s Global v2.1 Fundamental risk model. Each of these factors contributes towards the overall ex-ante volatility of the index, while underweights towards Volatility, Growth and Liquidity all help to reduce the overall risk profile of the index:

Stepping through into the supplier-expanded view of the FTSE 100, we see a very different set of risk sensitivities. Now the FTSE 100 gains a tilt towards both small cap and volatility and an underweight towards income:

Looking at the customer-expanded view of the FTSE 100, we see a similar tilt towards volatility, but an increase in the small cap exposure:

Conclusions

While they exist, it is becoming increasingly rare to find a company that has no connections to a wider network in some form. A firm’s customers are a vital source of revenue and therefore risk, while a company’s suppliers are a driver of revenue and therefore risk, as well.

While these network effects are often well understood at the individual stock level, they can be difficult to illustrate clearly in an intuitive way, particularly in the wider portfolio context. By incorporating supply chain data into our analysis, we can provide an additional lens through which to view a portfolio of stocks to more accurately visualize risk and sensitivities.