Hacker News Reader

Comment by hessenwolf | Hacker News Reader

You are throwing the baby out with the bath water, with respect to financial modelling. Yes, there are failures, and, yes, the models are severely imperfect.

We reduced the risk on a portfolio of 2 billion Euro, from a about a billion Euro to a risk of about 50 million Euro using hedging. The remaining 50 million was mostly basis risk, i.e., the mismatch between the underlying instruments in the liabilities and the hedge assets.

Using a similar logic to yours, senior management argued that we introduced a new risk called basis risk by trading derivatives.

View on HN

This might be naive of me, but how did you have a billion euro risk on a portfolio of 2 billion euros? How do you assess that risk to now be instead 50 million euros?

With respect, this seems like numbers were plucked out of somewhere. I propose that, if these come from a statistical model, it's the model that paints your position in the most optimistic light.

I don't judge you for that, but you should really read this article and look out for other fallacies of statistics.

There are other factors that affect the magnitude of risk that are independent of the model. One of them is leverage. Let's say, for simplicity's sake, there is a stock worth $100 and the assumption is that in a week it has 50% chance of becoming $90 and 50% chance of becoming $110. This assumption is your model and your risk is $10.

Now let's say you are bullish on the stock and borrow $400 to buy 5 stocks. In the same model, after gain or loss is realized and $400 is returned to the creditor after one week you are either $50 poorer or $50 richer, your magnitude of risk grew 5-fold by leveraging.

How do you model events such as yesterday's EUR/CHF?

Stochastic volatility and jump diffusion models

I think the Swiss Bank intervention is the classic black swan story. The model really only works until somebody puts their foot in it.

Risk management looks at the usual processes via stochastic modelling, but you need to also make scenario analysis, where you just make a deterministic event happen and then measure the effect on the value of your portfolio.

The risk wasn't 2 billion Euro, just the portfolio. The risk, according to solvency II, is 30-50% of the equities, approximately a 1% raise or drop in interest rates, 25% change in FX, etc.

The risk is probably about 40% of the 2 billion.

What is wrong with financial modeling, in my opinion, is not only that models cannot grasp too complex "reality", but that it is changing while you are finishing your model, so no statistical "snapshot" or data-set is even close to be correct.

Also, so-called Black swans could occur only within such models. There is no chance that one day c or even g could change (no matter what "scientists" used to say in journals).

Btw, finance is a business, not science.)

Yeah, so, I'm guessing you don't do a lot of financial modelling.

Not everyone is so lucky.

You don't actually wan't your model to be a fit to the data -- which I think you are implying it is.

> There is no chance that one day c or even g could change (no matter what "scientists" used to say in journals).

Well the thing is, c is defined in metres per second. And since 1983, the metre "has been defined as 'the length of the path travelled by light in vacuum during a time interval of 1/299,792,458 of a second.'".[0]

Since 1983, c is defined as constant.

[0] https://en.wikipedia.org/wiki/Metre

Reply on news.ycombinator.com