What the 2008 crisis teaches about decision models

Which public signals did a model have before the 2008 crisis?

The EQOS Asia back-test uses only data available in 2006 and 2007, drawn from public filings and market series, and it states that no privileged information was needed (EQOS Asia 2008 case page, as of October 2026). Housing prices, subprime share, derivative volumes, and bank leverage were all published. The first lesson for a model is that coverage of public sources already carries a large part of the warning.

A price-to-income ratio of 4.6 against a historical average of 3.2 shows how one ratio turns a vague boom into a measurable gap (EQOS Asia 2008 case page, as of October 2026). A decision model stores such ratios with their source and date, compares each to a reference, and lets the comparison feed a dimension score.

How does leverage show up in a 2008 decision model?

The EQOS Asia page lists bank leverage of about 30:1 in 2007 and computes that a fall of roughly 3.3% in asset values would erase the equity of such a bank (EQOS Asia 2008 case page, as of October 2026). The model reads this as a breaking threshold, a level beyond which the system tips, and it places the threshold beside the stress that actually occurred.

Stability is the dimension that carries the signal, and the page scores it at 0.08. A threshold converts a debate about risk appetite into a number that anyone can recompute. For a business decision today, the same logic applies to debt covenants, supplier concentration, or currency exposure: find the fall that ends the position, then compare it with plausible shocks.

Why does opacity matter for a decision model in 2008?

The page scores communication at 0.15 and argues that a large share of risk sat off the balance sheet, spread across layered securities that few holders could trace (EQOS Asia 2008 case page, as of October 2026). When exposure cannot be traced, the information that reaches decision makers carries more noise than signal.

The EQOS Asia decision engine expresses this through a signal-to-noise reading, which the page sets at X(S) = 0.185 for the system in 2006 and 2007. A low value tells the user that confidence in any other number should drop. A model that reports its own readability gives the reader an honest first answer: how far to trust the inputs.

Why did energy mislead observers during the 2008 boom?

Record volumes, profits, and product innovation made the system look healthy. The EQOS Asia page scores energy at 0.85 in the same period that integrity scores 0.08, and it describes that gap as the widest instability the method can measure (EQOS Asia 2008 case page, as of October 2026).

A simple average would rate the nine dimensions as middling and move on. The method on the same page combines dimensions through a weighted geometric product, so the weakest dimension pulls the total down. This design suits situations where one collapsed axis, such as integrity or care, can sink an otherwise energetic system.

What does the 2008 warning timeline say about decision models?

The page lists public events in sequence: HSBC reported subprime losses in February 2007, BNP Paribas froze three funds in August 2007, and Lehman Brothers filed for bankruptcy on 15 September 2008 (EQOS Asia 2008 case page, as of October 2026). Each event adds evidence, and a model that updates its inputs moves its score as the evidence arrives.

The back-test estimates a point of no return about 1.1 years after mid-2006, which places the tipping window between mid-2007 and early 2008 (EQOS Asia 2008 case page, as of October 2026). Read as a method check, the projection shows that a model can attach a time window to a score, and that the window can be compared with real dates afterward.

Which corrections would the 2008 model have recommended?

The page pairs the STOP signal with four corrections: consolidate off-balance-sheet exposure, restore a leverage ceiling of 15:1, separate raters from issuers, and verify repayment capacity before lending (EQOS Asia 2008 case page, as of October 2026). Each correction targets one weak dimension.

With the four corrections applied, the page recomputes φ(S) at 0.365 and describes a system in a work zone rather than a default zone (EQOS Asia 2008 case page, as of October 2026). The useful lesson for any model is constructive: a score comes with levers, and each lever has a target the user can monitor.

How can a reader apply the 2008 lessons to a decision today?

Start with the structure, not the headline. List the leverage, the opacity, and the safeguards in your own decision, then rate each dimension honestly. The simulator on the EQOS Asia site returns an estimate of φ(S) in about 10 minutes, free of charge (EQOS Asia case-study pages, as of October 2026).

For a full reading, the method page describes how the Audit pillar measures 9 dimensions and lists the 3 priority risks, and a 30-minute video call applies the equation to a real case. The 2008 study is retrospective, so it informs method choice and carries no investment advice.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile