Panel Reflection
Renee Fry spoke about financial crises and Prasanna Gai talked about modelling financial systems as complex networks. A key theme of this panel was links between players in the financial system, particularly the links between banks and the links between countries. Network diagrams and computer models of networks are key tools to analyse these complex systems. Another important tool is graphs, examples included graphs of exchange rates and share prices.
Complex issues raised during the panel included:
- How to avoid systematic risk?
- Should governments bail out banks?
- How should executive incentives work?
Some additional network concepts were introduced:
- percolation – how the network breaks into fragments
- contagion tends to spread with degrees of connectivity
Panel Question
I asked how useful/reliable are the network models of financial systems?
The models are not accurate at all and use a lot of judgement, however they do give rough rules of thumb that can be used to examine general policy impacts. You have to make a lot of assumptions about factors you can’t really predict, like assumptions about who is linked to who. Because there are not many historical financial crises there is a lack of historical information.
Tutorial Reflection
Our small group had to draw a network diagram of the GFC from the point of view of investors. This showed that there was too much separation between the physical asset and the investors, a theme already identified in the toolbox as “understand the underlying physical reality”.
Different people have different motivations for participating in complex systems. They may have conflicting goals and you can’t separate their motivations from the system. To understand the system you need to understand the goals of the people participating in the system.
Connections within this course
The lack of historical information about financial crises is the opposite problem to the excess data mentioned in the history panel.
The information about modelling financial systems as complex networks highlighted tools and concepts from other parts of this course.
- a network diagram of UK interbank network (with nodes and links), which is used to predict what happens if a bank fails [network diagrams]
- system stability and instability [engineering]
- similarities between financial systems and ecosystems [complex maths / fractals]
- some nodes or links are more important, they are the shocks that have the biggest impact [genetics seminar]
- risk sharing can become risk spreading
- financial systems, like all complex systems are ‘robust yet fragile’ because of their connectivity [ complex maths]
The reasons for Australia’s resilience during the global financial crisis can be considered in the context of the empires panel. If Kennedy’s thesis is correct then this financial resilience may have implications for military power. These reasons were:
- GDP growth stayed positive, most other countries went negative
- shock was external (in other countries) not internal
- Australia was in the middle of a commodities boom
- 4 pillars policy – main banks can’t take over each other
- higher interest rates than in other countries, therefore investors were less inclined to chase high-risk investments
- stimulus package – but was stimulus applied to the right areas? [lack of engineering]
Connections to other courses
The difficulty of predicting future events has been raised in Graphical Data Analysis and Applied Investments.
External Connections
Participants in financial crises may be categorised into:
- source – e.g. US housing market, currency market
- entities facilitating transmission of crisis
- countries/institutions with linkages
- countries/institutions without linkages
- investors
- ratings agencies – not fast enough to respond to change in risk
- policy makers: central banks (interest rates), regulators, international banks
- consumers: employment, superannuation, wages, real estate values
This method of categorisation and aggregation is, in some ways, what statistics is all about. For example, taking an average is a way of representing a large population with a single number. If it is done appropriately it allows complex problems to be simplified, because entities can be treated as groups rather than individually. The key is to identify which entities can be grouped and which are too dissimilar to be grouped.
Many financial models are done in spreadsheets, and their complexity can lead to errors (even “horror stories“). The spreadsheetrisks.com website has information about avoiding errors. Software engineering tools such as automatic testing and version management are difficult to apply to spreadsheets.
Tools to Address Complexity
- graphs and statistical methods
- network diagrams
- computer models
- categorisation and aggregation
- understand goals / motivations
So far in our tutorials we keep using network diagrams to analyse complex problems. In some ways this is satisfying since I introduced the concept to our tutorial group, but we are not getting other tools to address complex issues.