Hi all…my portfolio is made up of Mining assets split up into VC & PE. VC is exploration projects with some proven reserves (this includes: quarry, gold, diamonds etc) PE is growth capital for companies that need Finance.
 
The assets is 500Mn cash. Now I want to decide how much to allocate to the VC and I want to calculate this as the 1 year VAR. This means I am 95% confident that I will not lose more than this VAR and I commit this amount to the riskier portion of my portfolio, obviously with higher returns.
 
Conventionally you don’t take risk with VAR, you hold money to back it up as Capital.
 
Any takers?

The weakest link will be the model you use to estimate your loss distribution, not the allocation details.

How are you going to estimate the loss distribution?

As a general rule: you typically cannot have higher returns without higher risk when optimizing your portfolio.

Set your risk limit at whatever level of confidence you choose. Doesn't have to be 95%.
Construct some sort of distribution of returns (some people might use scenario analysis to do this).
Let's say that your risk limit is X million BWP and you know for the venture capital the most you will lose at the 95% level is Y% of your investment, then you would invest 100X/Y
If you want to, you can include the PE in the VaR calculation and include correlations between the PE and VC

Set your risk limit at whatever level of confidence you choose. Doesn't have to be 95%.
Construct some sort of distribution of returns (some people might use scenario analysis to do this).
Let's say that your risk limit is X million BWP and you know for the venture capital the most you will lose at the 95% level is Y% of your investment, then you would invest 100X/Y
If you want to, you can include the PE in the VaR calculation and include correlations between the PE and VC

Thanks Ppauper, you mean:
Set a notional risk limit say 10 million BWP, set my portfolio and calculate a 1 year VAR at a 95% CI say as 2 Million BWP i.e 20% of my invest. Then I decide to invest 10/0.2 = 50 Million BWP?

The weakest link will be the model you use to estimate your loss distribution, not the allocation details.

How are you going to estimate the loss distribution?

As a general rule: you typically cannot have higher returns without higher risk when optimizing your portfolio.

I want to invest an amount I am willing to lose but don't know how to derive this. Because VAR puts a limit to your losses at a given Confidence interval, I wanted the two to be linked. 

The weakest link will be the model you use to estimate your loss distribution, not the allocation details.

How are you going to estimate the loss distribution?

As a general rule: you typically cannot have higher returns without higher risk when optimizing your portfolio.

I want to invest an amount I am willing to lose but don't know how to derive this. Because VAR puts a limit to your losses at a given Confidence interval, I wanted the two to be linked. 

Yes, that makes sense. 
My worries is that you end up picking a popular scenario model -like the Geometric Brownian Motion that is used the the Black and Scholes world- and use that to estimate your VaR. It might say "in 95% of the cases the asset value won't drop more that $50mln" but my guess is that you won't be able to say something that precise with a mining operation and these commodity prices. Perhaps the model is wrong and a better model might say "ah, but if you include the risk of X,Y and Z happening -which are not *that* unlikely-, then the value can drop $400mln". In Texas X could be a hurricane, in Venezuela it might confiscation of property of banking issues. 

A hedge or an insurance against extreme losses would take away this model risk. The price of that insurance will of course be based on some model, but that's not your worry. If the price is right then you can eliminate your risk. If the price is very high then thats a sign that the insurance company thinks the risk is very high.

A hedge -if possible- is different from an insurance in the sense that it don't cost you money, it only changes the stability of your P&L. E.g. selling all future production at a fixed price is partial hedge. It's partial because there is always the fundamental risk of operational or legal issues that can stop the operation (in which case the hedge will be burden, you will have to deliver something you don't have).

Normally when you calculate VAR you would back it with money and hence have capital set aside in-case you lose up to the VAR, not take more risk with it, that's why I feel ppauper has a point, it's just that it's not easy to understand his model solution for now.

yes, but I hope it's clear that your money in not safe, A 95% VaR need be be breached 5% of the times for a model to be good. It can be 50% is your model is bad. 

Do you have some ideas about how to come up with a loss distribution? What losses can you expect with what probability?

Then model is not a problem...I aggregate at two levels:

• Line of business (& have correlations between Early Stage & Growth Capital)
• Commodities (& have correlations between Diamond, Coal, Quarries, Salt etc)
• The VAR is just multiplying correlation matrices with nominal values

...it is whether I can take risks with this VAR that is challenging. 

we're at cross-purposes it would seem.
I'll do a fairly simple example.
Suppose I have a portfolio of investments with a finite number of outcomes.
1% of the time, I lose everything.
3% of the time, I lose exactly 40% of my investment.
6% of the time, I lose exactly 25%
10% of the time, I lose exactly 10%
80% of the time, I gain exactly 15%

For this portfolio, the most you will lose at the 95% level is 25% of your investment.
If my risk limit is 10million BWP, then I would invest 40million BWP.
the most you will lose at the 95% level is 25% of your investment, and 25% of 40million is 10million, your risk limit

This doesn't answer your question and is not directly related but it might be something to keep in mind:
https://www.bloomberg.com/news/articles ... amond-debt

This doesn't answer your question and is not directly related but it might be something to keep in mind:
https://www.bloomberg.com/news/articles ... amond-debt

We are in a country with the best reputation for Mining and I don't see diamonds exceeding 20% of our portfolio.