Value at risk

In this episode of Twenty50 Investors, we investigate global markets, investment, and risk management.

From managing multi-asset portfolios to running a portfolio of businesses, we explore the age-old question: Does taking on greater risks truly result in higher return over time? What are the advantages and limits of stress tests, cross-asset correlation, value at risk, and heteroscedasticity? Finally, we explore whether the rise of machines and algorithmic trading could amplify systemic risks and lead to the next black swan.

And to further explore the complex world of risk management, today we are joined by Hatem Mustafa, co-head of global markets at Société Générale.

Hatem will share his unique insights in running a global markets platform and how to navigate the evolving world of risk management.

Not sure I've understood everything, but I get the idea.

Okay, let's finish with one last but important concept for risk management.

It's the value at risk, or VaR.

VaR? Is that a Nordic god with a hammer?

typical Wall Street risk management, a lot of focus is on what is the range of moderately probable outcomes.

People talk about VAR, value at risk, and lots of people try and build VAR systems.

What they are trying to is how wide is the band of outcomes, how wide is the band of stuff that has a 95% probability or a 98% probability, usually 95 or 98, but not 99.9.

If you know, I have a 98% chance of not losing a dollar.

That works in the VAR screen and the VAR, and it won't tell you, well, won't distinguish between whether the 2% chance is of losing $10 or losing $100.

To me, that's a very good that's a fundamentally wrong approach.

It's actually done because it's the sort of thing that statisticians are good at, To some degree, it's harder to imagine the worst thing than to statistically sort of figure out the middle band.

So we started the company with believing a lot that the data, and the impact that data can have, and how you model that will really transform different decision-making processes and the way people work.

But less we knew about LLMs, so, so we did a lot of forecasting of, you know, credit risk metrics, like probability of default, loss given default, prepayment rates, roll rates, transition matrices, pricing benchmark, a lot of VAR analysis on credit positions.

But never we thought of having, you know, the, the power that LLMs have given to us and what we are getting towards gen AI.

So I think the evolution of the last seven, eight years have been profound, and we all are seeing it, m- using it, trying it, seeing the value, and that's where I would say the market is today, Mark.

And that brings us to something I think boards should begin measuring explicitly.

It's called GVAR, geographic value at risk.

For every meaningful pipeline asset, score six variables.

One, innovation and IP concentration.

You don't need actuarial perfection.

You need visibility.

Because a 2 billion NPV asset with a geographic value at risk score of 27 should probably be discussed differently at the investment committee than another $2 billion asset scoring 11.

Now put a price on resilience.