Philip S. Douthit - Risk Allocation

Risk Allocation

Frequently Asked Questions

What is Risk Allocation? An answer to the basic question: How do I know what risks and how much of each risk to take

The key to a consistent framework that ties together all parts of the portfolio management process

How does it differ from traditional allocation practices? Risk Allocation approaches portfolio management from a different perspective

	Traditional Allocation		Risk Allocation
	Allocate % of assets		Allocate % of risk
	Focus on Returns		Focus on Risk Adjusted Returns
	Allocations not responsive to market conditions		Allocations change with changing market conditions
	Ex-post (historical returns)		Ex-ante (expected returns)
	Instrument based		Volatility and Correlation based<
	Constrained optimization		Eliminates corner solutions
	Risk measured after the fact		Risk measured before events occur
	Actual Risk varies widely with changing market conditions		Actual Risk held close to desired risk profile
	Reactive		Proactive
			Reconciles entrepreneur concentrations with portfolio diversification practice

Why is it better than traditional allocation practices? Because Risk Allocation balances risk among all the attractive opportunities, it: Allows each investment opportunity to contribute to returns Avoids surprises Better matches the investor's desired risk profile How do I know how much of which risks to take? In an ideal world you would have N opportunities which you liked equally and that had no correlation to each other. Logically you would spread the risk among them by giving each of them enough so that the risk of each was equal. But we don't live in an ideal world - you don't like the opportunities equally and they don't have zero correlations.

1/ Make an estimate of the expected return - and keep track of the estimates over time

This is not just an ex-post return calculation. This is an estimate based on all pertinent information (including historical performance) of what the opportunity will make going forward from the present.

2/ Look at the accuracy of the estimates (estimating error)

If you were a perfect estimator you would simply invest in the opportunity with the highest expected return. Note that you wouldn't care what the historical risk (volatility) is for that opportunity; because you were a perfect estimator you would know what the return will be and there is no risk.

But you are not a perfect estimator so as much as you like an opportunity you can't be sure and you start to diversify. " You may be a better estimator in one class of opportunities than in another class. While you would naturally favor that class, you would not want to totally exclude the other class. " You also try to spread your investments among desirable opportunities so that if something bad were to happen to one, the loss would only be proportional.

The problem is that most opportunities are linked and respond to varying degrees to the same external driving forces.

3/ Look at the correlations among the estimating errors

This is properly a study of the correlations between the errors of your estimates - not simply the correlation between the historical return performance.

4a/ you can use a conventional variance/covariance optimizer

Just remember to substitute your estimated returns for the 'returns', your estimating error for the 'volatility', and your estimating error correlations for the 'correlations'

4b/ or you can use Marginal Sharpe without resorting to the 'black box' of an optimizer

Marginal Sharpe = Relative Sharpe - Correlation
where Relative Sharpe is the ratio of the opportunity's Sharpe to the Portfolio's Sharpe, in both cases being your estimated return divided by your estimating error, and the correlation (?) is the correlation of your estimating error history for the opportunity to the aggregate of your estimating error histories for components of the portfolio

Simply stated, Marginal Sharpe says to add an opportunity until its correlation with the existing portfolio (including that opportunity) is equal to its Relative Sharpe.

The biggest lesson from Marginal Sharpe is that the correlation is the primary variable; as important as all the return estimates and estimating errors lumped together. And most people do not understand how to correctly calculate the correlation.

This is where we have to blend art with science. Because we may not be able to quantitatively measure our estimating error, much less our correlations, we may have to make reasonable estimates. But the estimates must be made in the context of understanding the true underlying mechanics outlined above. Only then will you have consistent estimates that are stable over time.

Even with the supporting data, the "devil is in the details". External events may not be reflected in systematic return estimates (i.e., elections). The correlation calculation may be dependant on the time frame used. Etc. How does it tie together the portfolio management process? Because all the components of portfolio management are derived from a shared perspective, they now fit naturally together.

Investment Policies start by defining the Investor's target risk level, and then go on to define how that risk can be diversified. Limits on how much risk for each asset class, key driver, country, sub-strategy, etc. are best defined before "the heat of battle".

Strategy and Manager Selection then qualify available opportunities for inclusion in the portfolio and determine their risk characteristics. Risk Based Allocation quantifies the amount of risk to allocate to each opportunity in order to meet the targets and limits of the Investment Policies.

Having allocated various risks, Portfolio Monitoring measures the actual risk incurred at the aggregate portfolio level, within each allocation, and for a variety of sub-groupings (by asset class, by country, by credit worthiness, etc.).

Because the portfolio manager has allocated risk, he now has a metric against which to measure the observed risk. Risk Compliance provides a definitive picture of whether the portfolio meets the objectives and limits set by the Investment Policies. Note that in addition to being excessive, the actual risk may also be below the desired target levels.

If the risk is outside the acceptable range, the portfolio manager now has the tools to act proactively - before actually incurring an excessive loss. Depending on the severity of a risk mismatch, the manger may tactically Intervene to solve an immediate problem, may routinely Rebalance to bring allocations back in line, or may strategically Reallocate to reflect changing market conditions.

With the information gathered around the cycle, the Manager and the Investor are now in a better position to review the Investment Policies and make changes reflecting changing political and economic conditions, changing market conditions (volatility), and new instruments and strategies.

The cycle is proactive; it acts to meet the Investor's objectives and then acts to correct any inadvertent deviations from the desired path.

Each step around the risk allocation cycle supports and depends on all other steps. Together they work toward the ultimate goal of maximizing the return while adhering to the Investor's risk profile.

Back or go to homepage