Would a model have predicted the 2008 crash? The EQOS back-test

Would a model have predicted the 2008 crash from public data?

According to the study, yes for the conditions and no for the date. The back-test uses only public data from 2006 and 2007: SEC filings, BIS statistics, FDIC reports and the S&P Case-Shiller index. It computes the same scores that EQOS Asia applies to clients today with the Codex φ v3.1 (EQOS Asia 2008 back-test, 2026).

The study frames the exercise as a retrospective educational analysis. Every input is verifiable and every computation is reproducible, so a reader can follow each step. The result is a reading of how fragile the system was, and the study leaves the exact timing of a crash open. That discipline separates a back-test from a story told after the fact, because the reader can rerun it.

Which housing and bank signals were public before the 2008 crash?

The study lists housing, derivatives and bank data that were available before the crisis. The price-to-income ratio stood at 4.6× against an average of 3.2×, the subprime share reached 23.5% against 8% in 2003, and subprime defaults hit 14.4% at the end of 2006. Also, 80% of CDO tranches carried an AAA rating. Each signal came from published sources, with no private information involved.

The page on the 2008 back-test data walks through those numbers one by one and shows why the study states that everything was visible in 2006 and 2007. Reading it first gives you the evidence base on which the score rests. 3 families of data matter: the real-estate market, the derivatives market and the bank balance sheets.

How do leverage and opacity show up in the 2008 back-test?

The study records bank leverage of 30:1 and notes that a fall of only 3.3% was enough to bankrupt a bank at that ratio. It also finds that 80% of risk sat off the balance sheet, so the real exposure was about 5 times the declared one (EQOS Asia 2008 back-test, 2026). For comparison, the study notes that the prudential threshold sat at 12 to 15:1 under Basel II.

Leverage and opacity form the core of the study’s reading, because they explain why a modest fall in asset prices became a system-wide event. The detailed page covers both ratios and the table of banks, with the threshold at which each would fail.

Which score did the back-test give the 2008 financial crisis system?

The study scores the global financial system at φ(S) = 0.234, below the critical threshold of 0.3, with 6 of 9 dimensions in the critical zone. Its ethical safeguard reads −0.187, which the study treats as an absolute block. Its signal-to-noise measure, X(S), reads 0.185, meaning more noise than signal. Each reading comes from the study and can be traced to the public data described on the detailed pages.

From that starting point the study projects a point of no return about 1.1 year after mid-2006. It then compares the projection with events: Bear Stearns sold for $2 a share in March 2008, against $172 a year before, and Lehman Brothers failed with $619 billion of debt on 15 September 2008. The study places those events within the window it projected.

What would EQOS have recommended in 2006 to lower the 2008 crash risk?

The study names 4 corrections: transparency, leverage, integrity and care. Transparency means putting off-balance-sheet exposures on the books. Leverage means restoring a limit of 15:1. Integrity means removing conflicts of interest at rating agencies. Care means checking a borrower’s ability to repay before lending. Each correction targets one of the dimensions that the study placed in the critical zone.

Applied together, the corrections lift the score to φ(S) = 0.365, above the 0.3 threshold. The study says the crisis would still have occurred, with a magnitude divided by 5 to 10. The page on what EQOS would have recommended in 2006 details each correction and the later reform that matched it.

Why did economic models see no risk before the 2008 crash?

The study contrasts its approach with DSGE models, the main tool of the IMF, the Fed and the ECB, which returned no systemic risk. It argues that these models assume rational agents, perfect information and linear behaviour, so ruptures cannot appear in them by construction. Because the model is linear, a fall in one variable produces a proportional reaction elsewhere, and a sudden break stays outside its vocabulary.

The page on why DSGE models saw no risk sets the 2 approaches side by side. It lists what a DSGE model leaves out and what a multi-dimension score measures instead, so a reader can judge where a model adds information and where it does not.

How should you use a back-test of the 2008 financial crisis?

Treat a back-test as a method check. It shows that a structured score applied to public data can flag fragility years ahead, and it invites you to run the same logic on your own case. This page is general information, not investment advice, and the study states that the scores do not predict dates. Run it first on a small, well-known case, then widen it as your confidence in the inputs grows.

The Λ simulator estimates a φ(S) score for your own system in minutes. For a decision in Singapore, ASEAN or Hong Kong, the EQOS Asia pre-diagnostic applies the scoring to your exposure, and the pillar guide on the stock market crash connects it with the history. Each session starts from your decision and ends with a ranked set of options.

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