Risk, we're told, is a latent variable. Therefore, no matter which risk measure we choose, it will always remain elusive, forever beyond our direct observation. We've heard this argument countless times. But is this correct?

Why there exists a “realized variance”, then? Why not a “realized Value at Risk”? What makes a risk measure observable or not? And what about the Expected Shortfall?

Let’s try to put some order.

We – risk managers – have all mistaken Value at Risk (VaR) forecasts for reality. We've confused model predictions with truth.

The most fundamental flaw of VaR isn't its lack of subadditivity or its inability to capture tail events, on which much has been written in the past, but its lack of observability. Have you ever noticed that "realized VaR" doesn't exist? Let's explore why.