Alpha in Academia

Alpha in Academia

The Diversification That Isn't There

[WITH CODE] What happens to portfolio risk when a single default hits your stocks and your bonds at the same moment.

Alpha in Academia's avatar
Alpha in Academia
Nov 15, 2026
∙ Paid

Hello and welcome back to another paid post!

Part 1 of 2. This piece builds the model and works out what it implies. Part 2 takes those implications to twenty years of market data and reports which of them survive, including the ones that do not.

On September 15, 2008, Lehman Brothers failed. Anyone holding Lehman paper lost it. But the S&P 500 also fell close to five percent that day, and that second loss landed on portfolios with no Lehman exposure whatsoever. Most risk systems are built to see those as two unrelated events. We built a model where they are the same event, and the numbers it produces are hard to look away from.

Let’s dive right in.


Introduction

Ask a risk manager how a balanced portfolio behaves in a crisis and you will usually get an answer assembled from parts. The equity book has a volatility, a beta, a factor structure. The credit book has spreads, ratings, default probabilities. Each is modelled well. The two are then stitched together with a correlation assumption, often a single number, sometimes stressed upward for bad times.

The stitching is where the problem lives. Treating equity risk and credit risk as two processes that happen to be correlated misses something structural about how crises actually arrive. A large default is not a credit event that coincidentally occurs during an equity selloff. It is frequently the thing that causes the equity selloff. The default and the drawdown are one event observed in two books.

The evidence is not subtle. Lehman’s failure in September 2008 took the Dow down roughly four and a half percent and the S&P down close to five percent in the same session. The collapse of Silicon Valley Bank in March 2023 was accompanied by heavy declines across banking equity indices in the United States and Europe, hitting institutions that held no SVB securities at all. In both cases a credit event propagated into equity prices immediately, discretely, and across names that had no direct exposure to the defaulted issuer.

This piece asks a narrow question with an uncomfortably wide answer. Suppose you hold a portfolio of stocks and a portfolio of defaultable bonds, and suppose the issuers of those bonds are entirely different companies from the ones whose stock you own. There is no overlap. On paper you are diversified across two asset classes and across dozens of distinct names. Now suppose that when one of those bond issuers defaults, two things happen at once: the bond goes to zero, and every stock you own takes a downward jump. How much risk are you actually carrying, and how badly does a conventional model understate it?

We built the model, validated it against analytical checks, and ran it. Three findings stand out. The first is a matter of magnitude: at a high level of default clustering, over a five-day horizon at a 99.9% confidence level, the portfolio loss our model produces is many times what an otherwise identical model without the shared default trigger produces. How many times depends heavily on how tightly defaults cluster, which is an input we choose here rather than measure. Part 2 measures it, and the answer turns out to be smaller than the figure this model produces at its most dramatic settings. The second is stranger and more useful. Raising the correlation between defaults does not reliably raise measured risk. Depending on which loss percentile you care about, it can push the number down. The third concerns the diversification claim itself. The two sleeves in our portfolio share no issuers and the model contains no direct correlation between them anywhere. The shared default trigger alone drives the correlation between them above 0.93.

Keep reading with a 7-day free trial

Subscribe to Alpha in Academia to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2026 Alpha in Academia · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for writing, podcasting, and media founders.