Asset modeling is so difficult

[CowboyGuy]

So far in my actuarial career, I had got to work on liabilities. This is the first time I'm getting to work on Assets (ALM).

I wonder if ALM is really ridiculously difficult or is it just me?

[MathGeek92]

not difficult at all

[Eimon Gnome]

Really complicated. A couple of things to keep in mind that may help:

1. Scenarios cannot be Risk Neutral
2. Open portfolios are much harder than closed block,a la Cash Flow Testing
3. In the end, you cannot solve for a best answer - there is only a risk/return trade off.

The first one should be clear. If you use RN scenarios, you will get back the answer that the best thing to do is cash flow match. Any other portfoio simply adds risk without adding return. This is not a particularly profound result and isn't much help.
So you need to have risk premia in the assets. The risk premia you assume drives your result. The premia could be excess returns on equities, excess returns from credit risk, or simply an arbitrage on the slope of the forward rate curve.

Open portfolios are always harder. It means you have to have forward risk premia. Suck it up and guess away.

Try and reduce your work to a risk/return proposition. An efficient frontier diagram is a common way to express the "equivalent" choices. Of course, you'll have to define risk in a numeric form. And that can lead to protracted philosophical discussions and endless stochastic runs.

Know that you are not alone. There is some comfort in that.

[CowboyGuy]

Originally Posted by Eimon Gnome

Really complicated. A couple of things to keep in mind that may help:

1. Scenarios cannot be Risk Neutral
2. Open portfolios are much harder than closed block,a la Cash Flow Testing
3. In the end, you cannot solve for a best answer - there is only a risk/return trade off.

The first one should be clear. If you use RN scenarios, you will get back the answer that the best thing to do is cash flow match. Any other portfoio simply adds risk without adding return. This is not a particularly profound result and isn't much help.
So you need to have risk premia in the assets. The risk premia you assume drives your result. The premia could be excess returns on equities, excess returns from credit risk, or simply an arbitrage on the slope of the forward rate curve.

Open portfolios are always harder. It means you have to have forward risk premia. Suck it up and guess away.

Try and reduce your work to a risk/return proposition. An efficient frontier diagram is a common way to express the "equivalent" choices. Of course, you'll have to define risk in a numeric form. And that can lead to protracted philosophical discussions and endless stochastic runs.

Know that you are not alone. There is some comfort in that.

That provides some comfort.

I have one more question here. The end results of liability modeling are mostly the profit margin, IRR, new business strain etc. I was wondering what is the output (result) of Asset modeling?

[Eimon Gnome]

Of the derived figures, the three I see time and time again are investment income, total return, and yield.

The yield typically drives a lot of the rate setting at an insurance firm. Projected yields translate into what interest credits managment sets. This appeals to the line managers and sales folks.

$ amount of investment income is the clearest link to the budget and the ytd income statment. This appeals to the CFO and probably your CEO.

While very few insurance companies spend much time on total return, it is a fairly standard performance measure at investment management firms such as mututal funds & hedge funds. I don't know much about the practice in pensions, so perhaps someone else can help there.

[MathGeek92]

didn't say it wasn't complicated... complicated and difficult are two separate things. IMO. I would suggest, take some time and learn all the pieces that go into asset modelling just like you did when you were learning how to model liabilities... credit spreads, default risks, prepayment risks, etc.

[Arthur Kade]

Originally Posted by Eimon Gnome

I don't know much about the practice in pensions, so perhaps someone else can help there.

[NoName]

Originally Posted by Arthur Kade

For the sake of those of us who are not pension people ... is the joke that pension asset modeling consists of "assume stocks always earn x% and bonds always earn y%"? Surely within the pension world there are people who do at least moderately sophisticated asset modeling?