Modern risk management practice has embraced data with unreserved enthusiasm. Risk management software providers compete based on the number of entries their databases contain. Only recently, one of the prominent firms in the space came out with a white paper on data-oriented systems, vaguely alluding to being “more accurate and precise” in their risk measurements. Incidentally, that firm is the record holder of the number of entries: over three million instruments. Since this number was not sufficient to help their clients survive the 2008 crisis, it would have to be augmented by the three million and first. Or would it?
While computers are very good at holding data in relational databases, they are rubbish at dealing with information. In fact, humans are hugely superior to computers when it comes to holding and manipulating information. Unfortunately, it is neither simple nor easy, and there is no guarantee of control of any sorts. Nevertheless, in the post-variance risk management world it is precisely this ability to work with information and not data that helps us draw conclusions and make decisions about the wild randomness in the markets. So how would we go about designing a risk process based on information?
Running a risk management function without identifying the dominant sources of global risk is akin running a military campaign without having a clue about who the adversary is. Global imbalances, excessive valuations or “bubbles” are a good place to start. This can be carried out on an ad-hoc basis, or methodically. The latter approach using Graham Risk measure is adopted by LINKS Analytics. A typical example of key global risk source is the infrastructure spending in China.
Once the biggest sources of global risk are identified, we can map the transmission pathways to other geographies, sectors and asset classes. The process is as simple as following the economic relationships between parties. A weakness in the Chinese infrastructure sector, for instance, would result in lower demand for materials, such as cold-rolled steel, cement, energy; municipal revenues from land sales will fall, which will put pressure on central government finances, etc. Quantifying these relationships is key to unlocking the potential risks to various asset classes driven by the network effects.
Estimating the impact of risk sources on asset returns is not difficult, since it is the direction of impact that is most informative and not the magnitude. Finally, although there are many imbalances in the global economy at any given point, only few of them pose an immediate threat. Therefore, the level of threat should be assessed dynamically. The ultimate goal of the risk process is to be prepared to implement dynamic hedging strategies or trigger de-risking in parts of the portfolio, should the level of threat be deemed sufficiently different from the past.
While the risk process described in this article is relatively simple and not too data intensive, it requires a risk function that is focused on the external market environment instead of reporting; and information and analysis instead of data. A risk analyst in this setting would require a qualification in business analysis and economics as opposed to SQL and quantitative methods.